English Syntax
English Syntax
English Syntax
. . . clause
. . .
. . . phrase
. . .
. . . word . . .
Typically we use the term clause to refer to a complete sentence-like unit, but which may be
part of another clause, as a subordinate or adverbial clause. Each of the sentences in (2b)(2d)
contains more than one clause, in particular, with one clause embedded inside another:
(2) a. The weather is lovely today.
b. I am hoping that [the weather is lovely today].
c. If [the weather is lovely today] then we will go out.
d. The birds are singing because [the weather is lovely today].
This chapter deals with what kind of combinatorial rules English employs in forming these
phrases, clauses, and sentences.
11
2.2 Lexical Categories
2.2.1 Determining the Lexical Categories
The basic units of syntax are words. The rst question is then what kinds of words (also known
as parts of speech, or lexical categories, or grammatical categories) does English have? Are
they simply noun, verb, adjective, adverb, preposition, and maybe a few others? Most of us
would not be able to come up with simple denitions to explain the categorization of words.
For instance, why do we categorize book as a noun, but kick as a verb? To make it more difcult,
how do we know that virtue is a noun, that without is a preposition, and that well is an adverb
(in one meaning)?
Words can be classied into different lexical categories according to three criteria: meaning,
morphological form, and syntactic function. Let us check what each of these criteria means,
and how reliable each one is.
At rst glance, it seems that words can be classied depending on their meaning. For exam-
ple, we could have the following rough semantic criteria for N (noun), V (verb), A (adjective),
and Adv (adverb):
(3) a. N: referring to an individual or entity
b. V: referring to an action
c. A: referring to a property
d. Adv: referring to the manner, location, time or frequency of an action
Though such semantic bases can be used for many words, these notional denitions leave a
great number of words unaccounted for. For example, words like sincerity, happiness, and pain
do not simply denote any individual or entity. Absence and loss are even harder cases.
There are also many words whose semantic properties do not match the lexical category that
they belong to. For example, words like assassination and construction may refer to an action
rather than an individual, but they are always nouns. Words like remain, bother, appear, and
exist are verbs, but do not involve any action.
A more reliable approach is to characterize words in terms of their forms and functions. The
form-based criteria look at the morphological form of the word in question:
(4) a. N: + plural morpheme -(e)s
b. N: + possessive s
c. V: + past tense -ed or 3rd singular -(e)s
d. V: + 3rd singular -(e)s
e. A: + -er/est (or more/most)
f. A: + -ly (to create an adverb)
According to these frames, where the word in question goes in the place indicated by , nouns
allow the plural marking sufx -(e)s to be attached, or the possessive s, whereas verbs can have
the past tense -ed or the 3rd singular form -(e)s. Adjectives can take comparative and superlative
12
endings -er or -est, or combine with the sufx -ly. (5) shows some examples derived from these
frames:
(5) a. N: trains, actors, rooms, mans, sisters, etc.
b. V: devoured, laughed, devours, laughs, etc.
c. A: fuller, fullest, more careful, most careful, etc.
d. Adv: fully, carefully, diligently, clearly, etc.
The morphological properties of each lexical category cannot be overridden; verbs cannot have
plural marking, nor can adjectives have tense marking. It turns out, however, that these morpho-
logical criteria are also only of limited value. In addition to nouns like information and furniture
that we presented in Chapter 1, there are many nouns such as love and pain that do not have
a plural form. There are adjectives (such as absent and circular) that do not have comparative
-er or superlative -est forms, due to their meanings. The morphological (form-based) criterion,
though reliable in many cases, is not a necessary and sufcient condition for determining the
type of lexical categories.
The most reliable criterion in judging the lexical category of a word is based on its syntactic
function or distributional possibilities. Let us try to determine what kind of lexical categories
can occur in the following environments:
(6) a. They have no .
b. They can .
c. They read the book.
d. He treats John very .
e. He walked right the wall.
The categories that can go in the blanks are N, V, A, Adv, and P (preposition). As can be seen
in the data in (7), roughly only one lexical category can appear in each position:
(7) a. They have no TV/car/information/friend.
b. They have no *went/*in/*old/*very/*and.
(8) a. They can sing/run/smile/stay/cry.
b. They can *happy/*down/*door/*very.
(9) a. They read the big/new/interesting/scientic book.
b. They read the *sing/*under/*very book.
(10) a. He treats John very nicely/badly/kindly.
b. He treats John very *kind/*shame/*under.
(11) a. He walked right into/on the wall.
b. He walked right *very/*happy/*the wall.
As shown here, only a restricted set of lexical categories can occur in each position; we can then
assign a specic lexical category to these elements:
13
(12) a. N: TV, car, information, friend, . . .
b. V: sing, run, smile, stay, cry, . . .
c. A: big, new, interesting, scientic, . . .
d. Adv: nicely, badly, kindly, . . .
e. P: in, into, on, under, over, . . .
In addition to these basic lexical categories, does English have other lexical categories? There
are a few more. Consider the following syntactic environments:
(13) a. student hits the ball.
b. John sang a song, Mary played the piano.
c. John thinks Bill is honest.
The only words that can occur in the open slot in (13a) are words like the, a, this, that, and so
forth, which are determiner (Det). (13b) provides a frame for conjunctions (Conj) such as and,
but, so, for, or, yet.
1
In (13c), we can have the category we call complementizer, here the word
that we return to these in (17) below.
Can we nd any supporting evidence for such lexical categorizations? It is not so difcult to
construct environments in which only these lexical elements appear. Consider the following:
(14) We found out that very lucrative jobs were in jeopardy.
Here we see that only words like the, my, his, some, these, those, and so forth can occur here.
These articles, possessives, quantiers, and demonstratives all determine the referential prop-
erties of jobs here, and for this reason, they are called determiners. One clear piece of evidence
for grouping these elements as the same category comes from the fact that they cannot occupy
the same position at the same time:
(15) a. *[My these jobs] are in jeopardy.
b. *[Some my jobs] are in jeopardy.
c. *[The his jobs] are in jeopardy.
Words like my and these or some and my cannot occur together, indicating that they compete
with each other for just one structural position.
Now consider the following examples:
(16) a. I think learning English is not easy at all.
b. I doubt you can help me in understanding this.
c. I am anxious you to study English grammar hard.
Once again, the possible words that can occur in the specic slot in (17) are strictly limited.
(17) a. I think that [learning English is not all that easy].
1
These conjunctions are coordinating conjunctions different from subordinating conjunctions like when, if, since,
though, and so forth. The former conjoins two identical phrasal elements whereas the latter introduces a subordinating
clause as in [Though students wanted to study English syntax], the department decided not to open that course this year.
14
b. I doubt if [you can help me in understanding this].
c. I am anxious for [you to study English grammar hard].
The italicized words here are different from the other lexical categories that we have seen so
far. They introduce a complement clause, marked above by the square brackets, and may be
sensitive to the tense of that clause. A tensed clause is known as a nite clause, as opposed to
an innitive. For example, that and if introduce or combine with a tensed sentence (present or
past tense), whereas for requires an innitival clause marked with to. We cannot disturb these
relationships:
(18) a. *I think that [learning English to be not all that easy].
b. *I doubt if [you to help me in understanding this].
c. *I am anxious for [you should study English grammar hard].
The term complement refers to an obligatory dependent clause or phrase relative to a head.
2
The italicized elements in (18) introduce a clausal complement and are consequently known as
complementizers (abbreviated as C). There are only a few complementizers in English (that,
for, if , and whether), but nevertheless they have their own lexical category.
Now consider the following environments:
(19) a. John not leave.
b. John drink beer last night.
c. John leave for Seoul tomorrow?
d. John will study syntax, and Mary , too.
The words that can appear in the blanks are neither main verbs nor adjectives, but rather words
like will, can, shall and must. In English, there is clear evidence that these verbs are different
from main verbs, and we call them auxiliary verbs (Aux). The auxiliary verb appears in front
of the main verb, which is typically in its citation form, which we call the base form. Note the
change in the main verb form in (20b) when the negation is added:
(20) a. He left.
b. He did not leave.
There is also one type of to which is auxiliary-like. Consider the examples in (21) and (22):
(21) a. Students wanted to write a letter.
b. Students intended to surprise the teacher.
(22) a. Students objected to the teacher.
b. Students sent letters to the teacher.
It is easy to see that in (22), to is a preposition. But how about the innitival marker to in (21),
followed by a base verb form? What lexical category does it belong to? Though the detailed
properties of auxiliary verbs will not be discussed until Chapter 8, we treat the innitival marker
2
See Chapter 4 for a fuller discussion of head and complement.
15
to as an auxiliary verb. For example, we can observe that to behaves like an auxiliary verb
should:
(23) a. It is crucial for John to show an interest.
b. It is crucial that John should show an interest.
(24) a. I know I should [go to the dentists], but I just dont want to.
b. I dont really want to [go to the dentists], but I know I should.
In (23), to and should introduce the clause and determines the tenseness of the clause. In (24),
they both can license the ellipsis of its VP complement.
3
Another property to shares with other auxiliary verbs like will is that it requires a base verb
to follow. Most auxiliary verbs are actually nite (tensed) forms which therefore pattern with
that in a nite clause, while the innitival clause introduced by for is only compatible with to:
(25) a. She thought it was likely [that everyone *to/might/would t into the car].
b. She thought it was easy [for everyone to/*might/*would t into the car].
Finally, there is one remaining category we need to consider, the particles (Part), illustrated
in (26):
(26) a. The umpire called off the game.
b. The two boys looked up the word.
Words like off and up here behave differently from prepositions, in that they can occur after the
object.
(27) a. The umpire called the game off .
b. The two boys looked the word up.
Such distributional possibilities cannot be observed with true prepositions:
(28) a. The umpire fell off the deck.
b. The two boys looked up the high stairs (from the oor).
(29) a. *The umpire fell the deck off .
b. *The students looked the high stairs up (from the oor).
We can also nd differences between particles and prepositions in combination with an object
pronoun:
(30) a. The umpire called it off . (particle)
b. *The umpire called off it.
(31) a. *The umpire fell it off .
b. The umpire fell off it. (preposition)
3
See Chapter 8 for detailed discussion on the ellipsis.
16
The pronoun it can naturally follow the preposition as in (31b), but not the particle in (30b).
Such contrasts between prepositions and particles give us ample reason to introduce another
lexical category Part (particle) which is differentiated from P (preposition). In the next section,
we will see more tests to differentiate these two types of word.
2.3 Grammar with Lexical Categories
As noted in Chapter 1, the main goal of syntax is building a grammar that can generate an
innite set of well-formed, grammatical English sentences. Let us see what kind of grammar
we can develop now that we have lexical categories. To start off, we will use the examples in
(32):
(32) a. A man kicked the ball.
b. A tall boy threw the ball.
c. The cat chased the long string.
d. The happy student played the piano.
Given only the lexical categories that we have identied so far, we can set up a grammar rule
for sentence (S) like the following:
(33) S Det (A) N V Det (A) N
The rule tells us what S can consist of: it must contain the items mentioned, except that those
which are in parentheses are optional. So this rule characterizes any sentence which consists of
a Det, N, V, Det, and N, in that order, possibly with an A in front of either N. We can represent
the core items in a tree structure as in (34):
(34)
S
.
.
.
.
.
.
.
.
.
.
.
.
.
Det N V Det N
. . . . . . . . . . . . . . .
We assume a lexicon, a list of categorized words, to be part of the grammar along with the rule
in (33):
(35) a. Det: a, that, the, this, . . .
b. N: ball, man, piano, string, student, . . .
c. V: kicked, hit, played, sang, threw, . . .
d. A: handsome, happy, kind, long, tall, . . .
By inserting lexical items into the appropriate pre-terminal nodes in the structure, where the
labels above . . . are, we can generate grammatical examples like those (32) as well as those like
the following, not all of which describe a possible real-world situation:
(36) a. That ball hit a student.
17
b. The piano played a song.
c. The piano kicked a student.
d. That ball sang a student.
Such examples are all syntactically well-formed, even if semantically in some cases, implying
that syntax is rather autonomous from semantics. Note that any anomalous example can be
preceded by the statement Now, heres something hard to imagine: . . . .
4
Notice that even this simple grammar rule can easily extend to generate an innite number
of English sentences by allowing iteration of the A:
5
(37) S Det A
N V Det A
N
The operator allows us to repeat any number of As, thereby generating sentences like (38). Note
that the parentheses around A in (34) are no longer necessary in this instance, for the Kleene
Star operator means any number including zero.
(38) a. The tall man kicked the ball.
b. The tall, handsome man kicked the ball.
c. The tall, kind, handsome man kicked the ball.
One could even generate a sentence like (39):
(39) The happy, happy, happy, happy, happy, happy man sang a song.
A grammar using only lexical categories can be specied to generate an innite number of
well-formed English sentences, but it nevertheless misses a great deal of basic properties that
we can observe. For example, this simple grammar cannot capture the agreement facts seen in
examples like the following:
(40) a. The mother of the boy and the girl is arriving soon.
b. The mother of the boy and the girl are arriving soon.
Why do the verbs in these two sentences have different agreement patterns? Our intuitions tell
us that the answer lies in two different possibilities for grouping the words:
(41) a. [The mother of [the boy and the girl]] is arriving soon.
b. [The mother of the boy] and [the girl] are arriving soon.
The different groupings shown by the brackets indicate who is arriving: in (41a), the mother,
while in (41b) it is both the mother and the girl. The grouping of words into larger phrasal units
which we call constituents provides the rst step in understanding the agreement facts in (41).
Now, consider the following examples:
(42) a. John saw the man with a telescope.
b. I like chocolate cakes and pies.
4
See Exercise 9 of this chapter and the discussion of selectional restrictions in Chapter 4.
5
This iteration operator
is called the Kleene Star Operator, and is a notation meaning zero to innitely many
occurrences. It should not be confused with the * prexed to a linguistic example, indicating ungrammaticality.
18
c. We need more intelligent leaders.
These sentences have different meanings depending on how we group the words. For example,
(42a) will have the following two different constituent structures:
(43) a. John saw [the man with a telescope].
(the man had the telescope)
b. John [[saw the man] with a telescope].
(John used the telescope)
Even these very cursory observations indicate that a grammar with only lexical categories is
not adequate for describing syntax. In addition, we need a notion of constituent, and need to
consider how phrases may be formed, grouping certain words together.
2.4 Phrasal Categories
In addition to the agreement and ambiguity facts, our intuitions may also lead us to hypothesize
constituency. If you were asked to group the words in (44) into phrases, what constituents would
you come up with?
(44) The student enjoyed his English syntax class last semester.
Perhaps most of us would intuitively assign the structure given in (45a), but not those in (45b)
or (45c):
(45) a. [The student] [enjoyed [his English syntax class last semester]].
b. [The] [student enjoyed] [his English syntax class] [last semester].
c. [The student] [[enjoyed his English] [syntax class last semester]].
What kind of knowledge, in addition to semantic coherence, forms the basis for our intuitions of
constituency? Are there clear syntactic or distributional tests which demonstrate the appropriate
grouping of words or specic constituencies? There are certain salient syntactic phenomena
which refer directly to constituents or phrases.
Cleft: The cleft construction, which places an emphasized or focused element in the X posi-
tion in the pattern It is/was X that . . . , can provide us with simple evidence for the existence
of phrasal units. For instance, think about how many different cleft sentences we can form from
(46).
(46) The policeman met several young students in the park last night.
With no difculty, we can cleft almost all the constituents we can get from the above sentence:
(47) a. It was [the policeman] that met several young students in the park last night.
b. It was [several young students] that the policeman met in the park last night.
c. It was [in the park] that the policeman met several young students last night.
d. It was [last night] that the policeman met several young students in the park.
19
However, we cannot cleft sequences that not form constituents:
6
(48) a. *It was [the policeman met] that several young students in the park last night.
b. *It was [several young students in] that the policeman met the park last night.
c. *It was [in the park last night] that the policeman met several young students.
Constituent Questions and Stand-Alone Test: Further support for the existence of phrasal
categories can be found in the answers to constituent questions, which involve a wh-word such
as who, where, when, how. For any given wh-question, the answer can either be a full sentence
or a fragment. This stand-alone fragment is a constituent:
(49) A: Where did the policeman meet several young students?
B: In the park.
(50) A: Who(m) did the policeman meet in the park?
B: Several young students.
This kind of test can be of use in determining constituents; we will illustrate with example (51):
(51) John put old books in the box.
Are either old books in the box or put old books in the box a constituent? Are there smaller
constituents? The wh-question tests can provide some answers:
(52) A: What did you put in your box?
B: Old books.
B: *Old books in the box.
(53) A: Where did you put the book?
B: In the box.
B: *Old books in the box.
(54) A: What did you do?
B: *Put old books.
B: *Put in the box.
B: Put old books in the box.
Overall, the tests here will show that old books and in the box are constituents, and that put old
books in the box is also a (larger) constituent.
The test is also sensitive to the difference between particles and prepositions. Consider the
similar-looking examples in (55), including looked and up:
(55) a. John looked up the inside of the chimney.
b. John looked up the meaning of chanson.
6
The verb phrase constituent met . . . night here cannot be clefted for independent reasons (see Chapter 12).
20
The examples differ, however, as to whether up forms a constituent with the following material
or not. We can again apply the wh-question test:
(56) A: What did he look up?
B: The inside of the chimney.
B: The meaning of chanson.
(57) A: Where did he look?
B: Up the inside of the chimney.
B: *Up the meaning of chanson.
(58) A: Up what did he look?
B: The inside of the chimney.
B: *The meaning of chanson.
What the contrasts here show is that up forms a constituent with the inside of the chimney in
(55a) whereas it does not with the meaning of chanson in (55b).
Substitution by a Pronoun: English, like most languages, has a system for referring back to
individuals or entities mentioned by the use of pronouns. For instance, the man who is standing
by the door in (59a) can be substituted by the pronoun he in (59b).
(59) a. What do you think the man who is standing by the door is doing now?
b. What do you think he is doing now?
There are other pronouns such as there, so, as, and which, which also refer back to other con-
stituents.
(60) a. Have you been [to Seoul]? I have never been there.
b. John might [go home], so might Bill.
c. John might [pass the exam], and as might Bill.
d. If John can [speak French uently] which we all know he can we will have no
problems.
A pronoun cannot be used to refer back to something that is not a constituent:
(61) a. John asked me to put the clothes in the cupboard, and to annoy him I really stuffed
them there [there=in the cupboard].
b. John asked me to put the clothes in the cupboard, and to annoy him I stuffed them
there [them=the clothes].
c. *John asked me to put the clothes in the cupboard, but I did so [=put the clothes] in
the suitcase.
Both the pronoun there and them refer to a constituent. However, so in (61c), referring to a VP,
refers only part of a constituent put the clothes, making it unacceptable.
21
Coordination: Another commonly-used test is coordination. Words and phrases can be co-
ordinated by conjunctions, and each conjunct is typically the same kind of constituent as the
other conjuncts:
(62) a. The girls [played in the water] and [swam under the bridge].
b. The children were neither [in their rooms] nor [on the porch].
c. She was [poor] but [quite happy].
d. Many people drink [beer] or [wine].
If we try to coordinate unlike constituents, the results are typically ungrammatical.
(63) a. *Mary waited [for the bus] and [to go home].
b. *Lee went [to the store] and [crazy].
Even though such syntactic constituent tests are limited in certain cases, they are often
adopted in determining the constituent of given expressions.
2.5 Phrase Structure Rules
We have seen evidence for the existence of phrasal categories. We say that phrases are projected
from lexical categories, and hence we have phrases such as NP, VP, PP, and so on. As before,
we use distributional evidence to classify each type, and then specify rules to account for the
distributions we have observed.
2.5.1 NP: Noun Phrase
Consider (64):
(64) [liked ice cream].
The expressions that can occur in the blank position here are once again limited. The kinds of
expression that do appear here include:
(65) Mary, I, you, students, the students, the tall students, the students from Seoul, the
students who came from Seoul, etc.
If we look into the sub-constituents of these expressions, we can see that each includes at least
an N and forms an NP (noun phrase). This leads us to posit the following rule:
7
(66) NP (Det) A* N (PP/S)
This rule characterizes a phrase, and is one instance of a phrase structure rule (PS rule). The rule
indicates that an NP can consist of an optional Det, any number of optional A, an obligatory N,
and then an optional PP or a modifying S.
8
The slash indicates different options for the same
place in the linear order. These options in the NP rule can be represented in a tree structure:
7
The relative clause who came from Seoul is kind of modifying sentence (S). See Chapter 11.
8
To license an example like the very tall man, we need to make A* as AP*. For simplicity, we just use the former in
the rule.
22
(67)
NP
.
.
.
.
.
.
.
.
.
.
.
.
.
(Det) A* N (PP/S)
. . . . . . . . . . . .
Once we insert appropriate expressions into the pre-terminal nodes, we will have well-formed
NPs; and the rule will not generate the following NPs:
(68) *the whistle tune, *the easily student, *the my dog, . . .
One important point is that as only N is obligatory in NP, a single noun such as Mary, you, or
students can constitute an NP by itself. Hence the subject of the sentence She sings will be an
NP, even though that NP consists only of a pronoun.
2.5.2 VP: Verb Phrase
Just as N projects an NP, V projects a VP. A simple test environment for VP is given in (69).
(69) The student .
(70) lists just a few of the possible phrases that can occur in the underlined position.
(70) snored, ran, sang, loved music, walked the dog through the park, lifted 50 pounds,
thought Tom is honest, warned us that storms were coming, etc.
These phrases all have a V as their head as projections of V, they form VP. VP can be charac-
terized by the rule in (71), to a rst level of analysis:
(71) VP V (NP) (PP/S)
This simple VP rule says that a VP can consist of an obligatory V followed by an optional NP
and then any number of PPs or an S. The rule thus does not generate ill-formed VPs such as
these:
(72) *leave the meeting sing, *the leave meeting, *leave on time the meeting, . . .
We can also observe that the presence of a VP is essential in forming a grammatical S, and the
VP must be nite (present or past tense). Consider the following examples:
(73) a. The monkey wants to leave the meeting.
b. *The monkey eager to leave the meeting.
(74) a. The monkeys approved of their leader.
b. *The monkeys proud of their leader.
(75) a. The men practice medicine.
b. *The men doctors of medicine.
These examples show us that an English well-formed sentence consists of an NP and a (nite)
VP, which can be represented as a PS rule:
23
(76) S NP VP
We thus have the rule that English sentences are composed of an NP and a VP, the precise
structural counterpart of the traditional ideas of a sentence being a subject and predicate or a
noun and a verb.
One more aspect to the structure of VP involves the presence of auxiliary verbs. Think of
continuations for the fragments in (77):
(77) a. The students .
b. The students want .
For example, the phrases in (78a) and (78b) can occur in (77a) whereas those in (78c) can appear
in (77b).
(78) a. run, feel happy, study English syntax, . . .
b. can run, will feel happy, must study English syntax, . . .
c. to run, to feel happy, to study English syntax, . . .
We have seen that the expressions in (78a) all form VPs, but how about those in (78b) and
(78c)? These are also VPs, which happen to contain more than one V. In fact, the parts after
the auxiliary verbs in (78b) and (78c) are themselves regular VPs. In the full grammar we
will consider to and can and so on as auxiliary verbs, with a feature specication [AUX +] to
distinguish them from regular verbs. Then all auxiliary verbs are simply introduced by a second
VP rule:
9
(79) VP V[AUX +] VP
One more important VP structure involves the VP modied by an adverb or a PP:
(80) a. John [[read the book] loudly].
b. The teacher [[met his students] in the class].
In such examples, the adverb loudly and the PP in the class are modifying the preceding VP. To
form such VPs, we need the PS rule in (81):
(81) VP VP Adv/PP
This rule, together with (76) will allow the following structure for (80b):
10
(82)
S
.
.
.
.
.
.
.
.
.
.
.
NP
VP
.
.
.
.
.
.
.
.
.
.
.
The teacher VP
PP
VP
.
.
.
.
.
.
.
.
.
.
.
.
Det A N V NP
PP
. . . . . . . . . . . . Det N P NP
. . . . . . . . . Det N
. . . . . .
With the structural possibilities shown here, let us assume that we have the following lexical
entries:
(100) a. Det: a, an, this, that, any, some, which, his, her, no, etc.
b. A: handsome, tall, fat, large, dirty, big, yellow, etc.
c. N: book, ball, hat, friend, dog, cat, man, woman, John, etc.
d. V: kicked, chased, sang, met, believed, thinks, imagines, assumes etc.
Inserting these elements in the appropriate pre-terminal nodes (the places with dots) in (99), we
are able to generate various sentences like those in (101):
13
(101) a. This handsome man chased a dog.
b. A man kicked that ball.
c. That tall woman chased a cat.
d. His friend kicked a ball.
There are several ways to generate an innite number of sentences with this kind of grammar.
As we have seen before, one simple way is to repeat a category (e.g., adjective) innitely. There
are also other ways of generating an innite number of grammatical sentences. Look at the
following two PS rules from (98) again:
13
The grammar still generates semantically anomalous examples like#The desk believed a man or#A man sang her
hat. For such semantically distorted examples, we need to refer to the notion of selectional restrictions (see Chapter
7).
27
(102) a. S NP VP
b. VP V S
As we show in the following tree structure, we can recursively apply the two rules, in the sense
that one can feed the other, and then vice versa:
(103)
S
7654 0123
NP VP
?>=< 89:;
N V S
7654 0123
John believes NP VP
?>=< 89:;
N V S
NP VP
7654 0123
N V[AUX +] VP
7654 0123
They will V NP
study English syntax
This means that we will also have a recursive structure like the following:
14
14
Due to the limited number of auxiliary verbs, and restrictions on their cooccurrence, the maximum number of
auxiliaries in a single English clause is 3. See Chapter 8.
28
(106)
S
.
.
.
.
.
.
.
.
.
NP VP
.
.
.
.
.
.
.
.
.
?>=< 89:;
N V [AUX +] VP
.
.
.
.
.
.
.
.
.
7654 0123
They will V[AUX +] VP
.
.
.
.
.
.
.
.
.
?>=< 89:;
have V[AUX +] VP
.
.
.
.
.
.
.
.
.
.
.
PP
V NP
Conj PP
Conj AP
Part
put the customers off
As represented here, the particle does not form a constituent with the following or preceding
NP whereas the preposition does form a constituent with it.
In summary, we have seen that a grammar with lexical categories can not only generate
an innite number of grammatical English sentences, but also account for some fundamental
properties, such as agreement and constituency.
17
This motivates the introduction of phrases
into the grammar.
17
In this chapter, we have not discussed the treatment of agreement with PS rules. Chapter 6 discusses the subject-
verb agreement in detail.
31
2.7 Exercises
1. Determine the lexical category of the italicized words in the following. In doing so, use
the three criteria (morphological, semantic, and syntactic) to provide the evidence for
your answer and state which one is the most reliable one.
(i) a. His second book came out earlier this year and became an instant best-
seller.
b. When you book something such as a hotel room, you arrange to have it.
c. Price quotes on selected categories will be sent out upon request.
d. No doubt that he was forced to leave his family against his will.
e. He intended to will the large amount of money to Frank.
f. Jane stood aside to let her pass.
g. He has a rail pass thats right for you.
h. It is important for us to spend time with children.
i. He was arrested for being drunk.
j. I think that person we met last week is insane.
k. We believe that he is quite reasonable.
l. I forgot to return the book that I borrowed from the teacher.
2. Consider the following data carefully and describe the similarities and differences among
that, for, if and whether. In so doing, rst compare that and for and then see how these
two are different from if and whether.
(i) a. I am anxious that you should arrive on time.
b. *I am anxious that you to arrive on time.
(ii) a. I am anxious for you to arrive on time.
b. *I am anxious for you should arrive on time.
(iii) a. I dont know whether/if I should agree.
b. I wonder whether/if youd be kind enough to give us information.
(vi) a. If students study hard, teachers will be happy.
b. Whether they say it or not, most teachers expect their students to study
hard.
3. Check if the italic parts form a constituent or not, using at least two constituenthood tests
(e.g., cleft, pronoun substitution, stand-alone, etc.). Also provide tree structures for each
of the following examples.
(i) a. John bought a book on the table.
b. John put a book on the table.
(ii) a. She turned down the side street
b. She turned down his offer.
(iii) a. He looked at a book about swimming.
b. He talked to a girl about swimming.
32
c. He talked with a girl about swimming.
(iv) a. I dont know the people present.
b. John called the president a fool.
4. Explain why the examples in (i) are ungrammatical. As part of the exercise, rst draw
structure for each example and try to determine the applicability of the the PS rules such
as the coordination rule in (110), presented earlier in this chapter.
(i) a. *Could you turn off the re and on the light?
b. *A nuclear explosion would wipe out plant life and out animal life.
c. *He ran down the road and down the President.
d. *I know the truth and that you are innocent.
e. *Lee went to the store and crazy.
5. Provide a tree structure for each of the following sentences and suggest what kind of VP
rules are necessary. In doing so, pay attention to the position of modiers like proudly,
by the park, and so forth.
(i) a. John refused the offer proudly.
b. I consider Telma the best candidate.
c. I saw him leaving the main building.
d. He took Masako to the school by the park.
e. John sang a song and danced to the music.
f. John wants to study linguistics in near future.
g. They told Angelica to arrive early for the award.
h. That Louise had abandoned the project surprised everyone.
6. Each of the following sentences is structurally ambiguous it has at least two structures.
Represent the structural ambiguities by providing different tree structures for each string
of words.
18
(i) a. I know you like the back of my hand.
b. I forgot how good beer tastes.
c. I saw that gas can explode.
d. Time ies like an arrow.
e. I need to have that report on our webpage by tomorrow.
7. Provide tree structures for each of the following sentences and see if there are any new
PS rules that we need to add, to supplement those we covered in this chapter. If there are
any places you cannot assign structures, please use triangles.
(i) Different languages may have different lexical categories, or they might
associate different properties to the same one. For example, Spanish uses
adjectives almost interchangeably as nouns while English cannot. Japanese
has two classes of adjectives whereas English has one; Korean, Japanese,
18
For i-e, to help tease out the ambiguity, consider related potential interpretations like Please put that book on my
desk. and That report on our webpage alleges that it does not function well..
33
and Chinese have measure words while European languages have nothing
resembling them; many languages dont have a distinction between adjec-
tives and adverbs, or adjectives and nouns, etc. Many linguists argue that
the formal distinctions between parts of speech must be made within the
framework of a specic language or language family, and should not be
carried over to other languages or language families.
19
19
Adapted from http://en.wikipedia.org/wiki/Part_of_speech
34
3
Syntactic Forms, Grammatical Functions, and
Semantic Roles
3.1 Introduction
In the previous chapter, we analyzed English sentences with PS rules. For example, the PS
rule S NP VP represents the basic rule for forming well-formed English sentences. As we
have seen, such PS rules allow us to represent the constituent structure of a given sentence in
terms of lexical and phrasal syntactic categories. There are other dimensions of the analysis
of sentences; one such way is using the notion of grammatical functions such as subject and
object:
(1) a. Syntactic categories: N, A, V, P, NP, VP, AP, . . .
b. Grammatical functions: SUBJ (Subject), OBJ (Object), MOD (Modier), PRED
(Predicate), . . .
The notions such as SUBJ, OBJ and PRED represent the grammatical function each constituent
plays in the given sentence. For example, consider one simple sentence:
(2) The monkey scratched a boy on Monday.
This sentence can be structurally represented in terms of either syntactic categories or gram-
matical functions as in the following:
(3) a. [
S
[
NP
The monkey] [
VP
scratched [
NP
a boy] [
PP
on Monday]]].
b. [
S
[
SUBJ
The monkey] [
PRED
scratched [
OBJ
a boy] [
MOD
on Monday]]].
As shown here, the monkey is an NP in terms of its syntactic form, but is the SUBJ (subject) in
terms of its grammatical function. The NP a boy is the OBJ (object) while the verb scratched
functions as a predicator. More importantly, we consider the entire VP to be a PRED (predicate)
which describes a property of the subject. On Monday is a PP in terms of its syntactic category,
but serves as a MOD (modier) here.
We also can represent sentence structure in terms of semantic roles. Constituents can be
considered in terms of conceptual notions of semantic roles such as agent, patient, location,
35
instrument, and the like. A semantic role denotes the underlying relationship that a participant
has with the relation of the clause, expressed by the main verb. Consider the semantic roles of
the NPs in the following two sentences:
1
(4) a. John tagged the monkey in the forest.
b. The monkey was tagged in the forest by John.
Both of these sentences describe a situation in which someone named John tagged a particular
monkey. In this situation, John is the agent and the monkey is the patient of the tagging event.
This in turn means that in both cases, John has the semantic role of agent (agt), whereas the
monkey has the semantic role of patient (pat), even though their grammatical functions are
different. We thus can assign the following semantic roles to each constituent of the examples:
(5) a. [[
agt
John] [
pred
tagged [
pat
the monkey] [
loc
in the forest]]].
b. [[
pat
The monkey] [
pred
was tagged [
loc
in the wood] [
agt
by John]]].
As noted here, in addition to agent and patient, we have predicate (pred) and locative (loc)
which also express the semantic role that each phrase performs in the described situation.
Throughout this book we will see that English grammar refers to these three different levels
of information (syntactic category, grammatical function, and semantic role), and they interact
with each other. For now, it may appear that they are equivalent classications: for example, an
agent is a subject and an NP, and a patient is an object and an NP. However, as we get further
into the details of the grammar, we will see many ways in which the three levels are not simply
co-extensive.
3.2 Grammatical Functions
How can we identify the grammatical function of a given constituent? Several tests can be used
to determine grammatical function, as we show here.
3.2.1 Subjects
Consider the following pair of examples:
(6) a. [The cat] [devoured [the rat]].
b. [The rat] [devoured [the cat]].
These two sentences have exactly the same words and have the same predicator devoured. Yet
they are signicantly different in meaning, and the main difference comes from what serves as
subject or object with respect to the predicator. In (6a), the subject is the cat, whereas in (6b)
it is the rat, and the object is the rat in (6a) but the cat in (6b).
The most common structure for a sentence seems to be one in which the NP subject is the one
who performs the action denoted by the verb (thus having the semantic role of agent). However,
this is not always so:
1
Semantic roles are also often called thematic roles or -roles (theta roles) in generative grammar (Chomsky
1982, 1986).
36
(7) a. My brother wears a green overcoat.
b. This car stinks.
c. It rains.
d. The committee disliked her proposal.
Wearing a green overcoat, stinking, raining, or disliking ones proposal are not agentive activi-
ties; they indicate stative descriptions or situations. Such facts show that we cannot rely on the
semantic roles of agent for determining subjecthood.
More reliable tests for subjecthood come from syntactic tests such as agreement, tag ques-
tions, and subject-auxiliary inversion.
Agreement: The main verb of a sentence agrees with the subject in English:
(8) a. She never writes/*write home.
b. These books *saddens/sadden me.
c. Our neighbor takes/*take his children to school in his car.
As we noted in Chapter 1, simply being closer to the main verb does not entail subjecthood:
(9) a. The book, including all the chapters in the rst section, is/*are very interesting.
b. The effectiveness of teaching and learning *depend/depends on several factors.
c. The tornadoes that tear through this county every spring *is/are more than just a
nuisance.
The subject in each example is book, effectiveness, and tornadoes respectively, even though
there are nouns closer to the main verb. This indicates that it is not simply the linear position of
the NP that determines agreement; rather, agreement shows us what the subject of the sentence
is.
Tag questions: A tag question, a short question tagged onto the end of an utterance, is also a
reliable subjecthood test:
(10) a. The lady singing with a boy is a genius, isnt she/*isnt he?
b. With their teacher, the kids have arrived safely, havent they/ *hasnt he?
The pronoun in the tag question agrees with the subject in person, number, and gender it
refers back to the subject, but not necessarily to the closest NP, nor to the most topical one. The
pronoun she in (10a) shows us that lady is the head of the subject NP in that example, and they
in (10b) leads us to assign the same property to kids. The generalization is that a tag question
must contain a pronoun which identies the subject of the clause to which the tag is attached.
Subject-auxiliary inversion: In forming questions and other sentence-types, English has
subject-auxiliary inversion, which applies only to the subject.
(11) a. This teacher is a genius.
b. The kids have arrived safely.
37
c. It could be more detrimental.
(12) a. Is this teacher a genius?
b. Have the kids arrived safely?
c. Could it be more detrimental?
As seen here, the formation of Yes/No questions such as these involves the rst tensed auxil-
iary verb moving across the subject: more formally, the auxiliary verb is inverted with respect to
the subject, hence the term subject-auxiliary inversion. This is not possible with a non-subject:
(13) a. The kids in our class have arrived safely.
b. *Have in our class the kids arrived safely?
Subject-auxiliary inversion provides another reliable subjecthood test.
3.2.2 Direct and Indirect Objects
A direct object (DO) is canonically an NP, undergoing the process denoted by the verb:
(14) a. His girlfriend bought this computer.
b. That silly fool broke the teapot.
However, this is not a solid generalization. The objects (OBJ) in (15a) and (15b) are not really
affected by the action. In (15a) the dog is experiencing something, and in (15b) the thunder is
somehow causing some feeling in the dog:
(15) a. Thunder frightens [the dog].
b. The dog fears [thunder].
Once again, the data show us that we cannot identify the object based on semantic roles. A
much more rm criterion is the syntactic construction of passivization, in which a notional
direct object appears as subject. The sentences in (16) can be turned into passive sentences in
(17):
(16) a. His girlfriend bought this computer for him.
b. The child broke the teapot by accident.
(17) a. This computer was bought for him by his girlfriend.
b. The teapot was broken by the child by accident.
What we can notice here is that the objects in (16) are promoted to subject in the passive
sentences. The test comes from the fact that non-object NPs cannot be promoted to the subject:
(18) a. This item belongs to the student.
b. *The student is belonged to by this item.
(19) a. He remained a good friend to me.
b. *A good friend is remained to me (by him).
38
The objects that undergo passivization are direct objects, distinct from indirect objects.
An indirect object (IO) is one which precedes a direct object (DO), as in (20); IOs are NPs
and have the semantic roles of goal, recipient, or benefactive:
(20) a. I threw [the puppy] [the ball]. (IO = goal)
b. John gave [the boys] [the CDs]. (IO = recipient)
c. My mother baked [me] [a birthday cake]. (IO = benefactive)
In examples like (20), passive has the property of making the IO into the subject.
(21) a. The boys were given the CDs (by John).
b. She was sent a review copy of the book (by the publisher).
Note that sentences with the IO-DO order are different from those where the semantic role
of the IO is expressed as an oblique PP, following the DO (such a process is sometimes called
dative alteration):
(22) a. John gave the CDs to the boys.
b. The publisher sent a review copy of the book to her.
c. My mother baked a cake for me.
In this kind of example, it is once again the DO which can be passivized, giving examples like
the following:
(23) a. The CDs were given to the boys by John.
b. A review copy of the book was sent to her by the publisher.
c. This nice cake was baked for me by my mother.
3.2.3 Predicative Complements
There also are NPs which follow a verb but which do not behave as DOs or IOs. Consider the
following sentences:
(24) a. This is my ultimate goal.
b. Michelle became an architect.
(25) a. They elected Graham chairman.
b. I consider Andrew the best writer
The italicized elements here are traditionally called predicative complements in the sense that
they function as the predicate of the subject or the object. However, even though they are NPs,
they do not passivize:
(26) a. *Chairman was elected Graham.
b. *The best writer was considered Andrew.
The difference between objects and predicative complements can also be seen in the following
contrast:
(27) a. John made Kim a great doll.
39
b. John made Kim a great doctor.
Even though the italicized expressions here are both NPs, they function differently. The NP a
great doll in (27a) is the direct object, as in John made a great doll for Kim, whereas the NP
a great doctor in (27b) cannot be an object: it serves as the predicate of the object Kim. If we
think of part of the meaning informally, only in the second example would we say that the nal
NP describes the NP Kim.
(28) a. (27)a: Kim = a great doll
b. (27)b: Kim = a great doctor
In addition, phrases other than NPs can serve as predicative complements:
(29) a. The situation became terrible.
b. This map is what he wants.
c. The message was that you should come on time.
(30) a. I made Kim angry.
b. I consider him immoral.
c. I regard Andrew as the best writer.
d. They spoil their kids rotten.
The italicized complements function to predicate a property of the subject in (29) and of the
object in (30).
3.2.4 Oblique Complements
Consider now the italicized expressions in (31):
(31) a. John put books in the box.
b. John talked to Bill about the exam.
c. She reminded him of the last time they met.
d. They would inform Mary of any success they have made.
These italicized expressions are neither objects nor predicative complements. Since their pres-
ence is obligatory, for syntactic well-formedness, they are called oblique complements. Roughly
speaking, oblique contrasts with the direct functions of subject and object, and oblique
phrases are typically expressed as PPs in English.
As we have seen before, most ditransitive verbs can also take oblique complements:
(32) a. John gave a book to the student.
b. John bought a book for the student.
The PPs here, which cannot be objects since they are not NPs, also do not serve as predicate of
the subject or object they relate directly to the verb, as oblique complements.
40
3.2.5 Modiers
The functions of DO, IO, predicative complement, and oblique complement all have one com-
mon property: they are all selected by the verb, and we view them as being present to comple-
ment the verb to form a legitimate VP. Hence, these are called complements (COMPS), and
typically they cannot be omitted.
Unlike these COMPS, there are expressions which do not complement the predicate in the
same way, and which are truly optional:
(33) a. The bus stopped suddenly.
b. Shakespeare wrote his plays a long time ago.
c. They went to the theater in London.
d. He failed chemistry because he cant understand it.
The italicized expressions here are all optional and function as modiers (also called adjuncts
or adverbial expressions). These modiers specify the manner, location, time, or reason,
among many other properties, of the situations expressed by the given sentences informally,
they are the (how, when, where, and why) phrases.
One additional characteristic of modiers is that they can be stacked up, whereas comple-
ments cannot.
(34) a. *John gave Tom [a book] [a record].
b. I saw this lm [several times] [last year] [during the summer].
As shown here, temporal adjuncts like several times and last year can be repeated, whereas the
two complements a book and a record in (34a) cannot. Of course, temporal adjuncts do not
become the subject of a passive sentence, suggesting that they cannot serve as objects.
(35) a. My uncle visited today.
b. *Today was visited by my uncle.
3.3 Form and Function Together
We now can analyse each sentence in terms of grammatical functions as well as the structural
constituents. Let us see how we can analyze a simple sentence along these two dimensions:
(36)
S
_
_
_
_
_
_
_
_
_
_
_
_
_
_
NP: SUBJ
VP: PRED
N V CP: OBJ
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
John believes C S
.
.
.
.
.
.
.
.
.
.
.
.
VP: PRED
VP: PRED
_
_
_
_
_
_
_
_
_
_
_
_
_
_
Det A N VP: PRED
NP: MOD
last night
devoured a mouse
Here the expression last night is an adverbial NP in the sense that it is categorically an NP but
functions as a modier (adjunct) to the VP.
As we go through this book, we will see that the distinction between grammatical functions
and categorical types is crucial in the understanding of English syntax.
3.4 Semantic Roles
As noted before, semantic roles were introduced as a way of classifying the arguments of pred-
icators (mostly verbs and adjectives) into a closed set of participant types. Even though we
43
cannot make any absolute generalizations about the relationship between grammatical func-
tions and semantic roles, the properties of semantic roles do interact in regular ways with cer-
tain grammatical constructions. A list of the most relevant thematic roles and their associated
properties is given below.
5
Agent: A participant which the meaning of the verb species as doing or causing some-
thing, possibly intentionally. Examples: subject of eat, kick, hit, hammer, etc.
(44) a. John ate his noodle quietly.
b. A boy hit the ball.
c. A smith hammered the metal.
Patient: A participant which the verb characterizes as having something happen to it, and
as being affected by what happens to it. Examples: object of kick, hit, hammer, etc.
6
(45) a. A boy hit the ball.
b. A smith hammered the metal.
Experiencer: A participant who is characterized as aware of something. Examples: sub-
ject of perception verbs like feel, smell, hear, see, etc.
(46) a. The students felt comfortable in the class.
b. The student heard a strange sound.
Theme: A participant which is characterized as changing its position or condition, or
as being in a state or position. Examples: direct object of give, hand, subject of come,
happen, die, etc.
(47) a. John gave a book to the students.
b. John died last night.
Benefactive: The entity that benets from the action or event denoted by the predicator.
Examples: oblique complement of make, buy, etc.
(48) a. John made a doll for his son.
b. John bought a lot of books for his sons.
Source: The one from which motion proceeds. Examples: subject of promise, object of
deprive, free, cure, etc.
(49) a. John promised Bill to leave tomorrow morning.
b. John deprived his sons of game cards.
Goal: The one to which motion proceeds. Examples: subject of receive, buy, indirect
object of tell, give, etc.
(50) a. Mary received an award from the department.
b. John told the rumor to his friend.
5
The denition of semantic roles given here is adopted from Dowty (1989).
6
Patient and theme are often unied into undergoer in the sense that both the patient and theme individual undergo
the action in question.
44
Location: The thematic role associated with the NP expressing the location in a sentence
with a verb of location. Examples: subject of keep, own, retain, locative PPs, etc.
(51) a. John put his books in the attic.
b. The government kept all the money.
Instrument: The medium by which the action or event denoted by the predicator is carried
out. Examples: oblique complement of hit, wipe, hammer, etc.
(52) a. John hit the ball with a bat.
b. John wiped the window with a towel.
An important advantage of having such semantic roles available to us is that it allows us to
capture the relationship between two related sentences, as we have already seen. As another
example, consider the following pair:
(53) a. [
agt
The cat] chased [
pat
the mouse].
b. [
pat
The mouse] was chased by [
agt
the cat].
Even though the above two sentences have different syntactic structures, they have essentially
identical interpretations. The reason is that the same semantic roles assigned to the NPs: in both
examples, the cat is the agent, and the mouse is the patient. Different grammatical uses of verbs
may express the same semantic roles in different arrays.
The semantic roles also allow us to classify verbs into more ne-grained groups. For exam-
ple, consider the following examples:
(54) a. There still remains an issue to be solved.
b. There lived a man with his grandson.
c. There arrived a tall, red-haired and incredibly well dressed man.
(55) a. *There sang a man with a pipe.
b. *There dances a man with an umbrella.
All the verbs are intransitive, but not all are acceptable in the there-construction. The dif-
ference can come from the semantic role of the postverbal NP, as assigned by the main verb.
Verbs like arrive, remain, live are taken to assign the semantic role of theme (see the list of
roles above), whereas verbs like sing, dance assign an agent role. We thus can conjecture that
there-constructions do not accept the verb whose subject carries an agent semantic role.
While semantic roles provide very useful ways of describing properties across different con-
structions, we should point out that the theoretical status of semantic roles is still unresolved.
7
For example, there is no agreement about exactly which and how many semantic roles are
needed. The problem is illustrated by the following simple examples:
(56) a. John resembles his mother.
b. A is similar to B.
7
See Levin and Rappaport Hovav (2005) for further discussion of this issue.
45
What kind of semantic roles do the arguments here have? Both participants seem to be playing
the same role in these examples they both cannot be either agent or patient or theme. They are
also cases where we might not be able to pin down the exact semantic role:
(57) a. John runs into the house.
b. Mary looked at the sky.
The subject John in (57a) is both agent and theme: it is agent since it initiates and sustains the
movement but also theme since it is the object that moves.
8
Also, the subject Mary in (57b) can
either be an experiencer or an agent depending on her intention one can just look at the sky
with no purpose at all.
9
Even though there are theoretical issues involved in adopting semantic roles in the grammar,
there are also many advantages of using them. We can make generalizations about the grammar
of the language: typically the agent takes the subject position, while an NP following the
word from is serving as the source. As we will see in the next chapter, semantic roles are
also recognized as the standard concepts used for organizing predicate-argument structures for
predicates within the lexicon. In the subsequent chapters, we will refer to semantic roles in
various places.
3.5 Exercises
1. Construct sentences containing the following grammatical functions:
(i) a. subject, predicator, direct object
b. subject, predicator, indirect object, direct object
c. subject, predicator, adjunct
d. adjunct, subject, predicator
e. adjunct, subject, predicator, direct object
f. subject, predicator, direct object, oblique complement
g. subject, predicator, predicative complement
h. subject, predicator, direct object, predicative complement
i. subject, predicator, predicative complement, adjunct
j. subject, predicator, direct object, predicative complement, adjunct
2. Give the grammatical function of the italicized phrases in the following examples:
(i) a. All of his conversation was reported to me.
8
Jackendoff (1987) develops an account of thematic roles in which agency and motion are two separate dimensions,
so, in fact, a single NP can be agent and theme.
9
To overcome the problemof assigning the correct semantic role to an argument, one can assume that each predicator
has its own (individual) semantic roles. For example, the verb kick, instead of having an agent and a patient, has two
individualized semantic roles kicker and kicked. See Pollard and Sag (1987).
46
b. Sandy removed her ballet shoes.
c. The school awarded a few of the girls in Miss Kims class scholarships.
d. She was the nicest teacher in the Senior School.
e. They elected him Americas 31st President.
f. The next morning we set out for Seoul.
g. Doing syntax is not easy.
h. This is the place to go to.
i. He saw the man with the stick.
j. This week will be a difcult one for us.
k. We need to nish the project this week.
3. Draw tree structures for the following sentences and then assign an appropriate gram-
matical function to each phrase.
(i) a. They parted the best of friends.
b. In the summer we always go to France.
c. Benny worked in a shoe factory when he was a student.
d. Last year I saw this lm several times.
e. He baked Tom the bread last night.
f. That they have completed the course is amazing.
g. Everyone hoped that she would sing.
h. The gang robbed her of her necklace.
i. They helped us edit the script.
j. The teacher made students happy.
k. We reminded him of the agreement.
4. Consider the following examples:
(i) a. There is/*are only one chemical substance involved in nerve transmission.
b. There *is/are more chemical substances involved in nerve transmission.
With respect to the grammatical function of there, what can we infer from these data? Try
out more subjecthood tests such as the tag-question test to determine the grammatical
function of there in these examples. In addition, try to decide what is the subject in
the following so-called locative inversion examples and provide at least three different
locative inversion examples that you can nd from naturally-occurring material.
(ii) a. In the garden stands/*stand a statue.
b. Among the guests was/*were sitting my friend Louise.
5. Determine the grammatical function of the italiczed phrase, providing at least one syn-
tactic test we have discussed in the chapter.
10
(i) a. This proved a decisive factor.
b. This proved my hypothesis.
10
This exercise is adopted from Huddleston and Pullum (2002).
47
c. The students all enjoyed that summer.
d. The students all worked that summer.
e. The scientist made her a robot.
f. The students called me a teacher.
6. Assign a semantic role to each argument in the following sentences.
(i) a. A big green insect ew into the soup.
b. Johns mother sent a letter to Mary.
c. John smelled the freshly baked bread.
d. We placed the cheese in the refrigerator.
e. Frank threw himself into the sofa.
f. The crocodile devoured the doughnut.
g. John came from Seoul.
h. John is afraid of Bill.
i. The ice melted.
j. The vacuum cleaner terries the child.
7. Determine the grammatical functions for the underlined expressions in the following
text.
(i) Scientists found that the birds sang well in the evenings, but performed
badly in the mornings. After being awake several hours, however, the
young males regained their mastery of the material and then improved on
the previous days accomplishments. To see whether this dip in learning
was caused by the same kind of pre-coffee fog that many people feel in
the morning, the researchers prevented the birds from practicing rst thing
in the morning. They also tried keeping the birds from singing during the
day, and they used a chemical called melatonin to make the birds nap at
odd times. The researchers concluded that their study supports the idea
that sleep helps birds learn. Studies of other animals have also suggested
that sleep improves learning.
11
11
From Science News Online, Feb 2, 2007
48
4
Head, Complements, and Modiers
4.1 Projections from Lexical Heads to Phrases
4.1.1 Internal vs. External Syntax
As we have seen in the previous chapters, both syntactic categories (NP, AP, VP, PP, etc.) and
grammatical functions (subject, complement, and modier) play important roles in the analysis
of English sentences. We have also observed that the grammatical function and form of each
constituent depend on where it occurs or what it combines with.
The combinatory properties of words and phrases involve two aspects of syntax: internal
and external syntax.
1
Internal syntax deals with how a given phrase itself is constructed in a
well-formed manner whereas external syntax is concerned with how a phrase can be used in a
larger construction. Observe the following examples:
(1) a. *John [put his gold].
b. *John [put under the bathtub].
c. *John [put his gold safe].
d. *John [put his gold to be under the bathtub].
e. John [put his gold under the bathtub].
Why is only (1e) acceptable? Simply, because only it satises the condition that the verb put
selects an NP and a PP as its complements, and it combines with them in the VP. In the other
examples, this condition is not fullled. This combinatory requirement starts from the internal
(or lexical) properties of the verb put, and is not related to any external properties of the VP.
In contrast, the external syntax is concerned with the external environment in which a phrase
occurs. Some of the unacceptable examples in (1) can be legitimate expressions if they occur in
the proper (syntactic) context.
(2) a. This is the box in which John [put his gold]. (cf. (1a))
b. This is the gold that John [put under the bathtub]. (cf. (1b))
1
The terms internal and external syntax are from Baker (1995).
49
Meanwhile, the well-formed VP in (1e) can be unacceptable, depending on external contexts.
For example, consider frame induced by the governing verb kept in (3):
(3) a. *The king kept [put his gold under the bathtub].
b. The king kept [putting his gold under the bathtub].
The VP put his gold under the bathtub is a well-formed phrase, but cannot occur in (3a) since
this is not the environment where such a nite VP occurs. That is, the verb kept requires the
presence of a gerundive VP like putting his gold under the bathtub, and therefore imposes an
external constraint on VPs.
4.1.2 Notion of Head, Complements, and Modiers
One important property we observe in English phrase-internal syntax is that in building up any
phrase, there is one obligatory element in each phrase. That is, each phrase has one essential
element as represented in the diagrams in (4):
(4) a.
NP
. . . N
7654 0123
b.
VP
V
7654 0123
. . .
c.
AP
A
7654 0123
. . .
The circled element here is the essential, obligatory element within the given phrase. We call
this essential element the head of the phrase.
2
The head of each phrase thus determines its
projection into a larger phrasal constituent. The head of an NP is thus N, the head of a VP is
V, and the head of an AP is A.
The notion of headedness plays an important role in the grammar. For example, the verb
put, functioning as the head of a VP, dictates what it must combine with two complements,
NP and PP. Consider other examples:
(5) a. The defendant denied the accusation.
b. *The defendant denied.
(6) a. The teacher handed the student a book.
b. *The teacher handed the student.
The verb denied here requires an NP object whereas handed requires two NP complements, in
this use. The properties of the head verb itself determine what kind of elements it will combine
with. As noted in the previous chapter, the elements which a head verb should combine with
are called complements. The complements include direct object, indirect object, predicative
complement, and oblique complement since these are all potentially required by some verb or
other.
The properties of the head become properties of the whole phrase. Why are the examples in
(7b) and (8b) ungrammatical?
2
See section 1.3 in Chapter 1 also.
50
(7) a. They [want to leave the meeting].
b. *They [eager to leave the meeting].
(8) a. The senators [know that the president is telling a lie].
b. *The senators [certain that the president is telling a lie].
The examples in (7b) and (8b) are unacceptable because of the absence of the required head.
The unacceptable examples lack a nite (tensed) VP as the bracketed part, but we know that
English sentences require a nite VP as their immediate constituent, as informally represented
as in (9):
(9) English Declarative Sentence Rule:
Each declarative sentence must contain a nite VP as its head.
Each nite VP is headed by a nite verb. If we amend the ungrammatical examples above to
include a verb but not a nite one, they are still ungrammatical:
(10) a. *They [(to) be eager to leave the meeting].
b. *The senators [(to) be certain that the president is telling a lie].
The VP is considered to be the (immediate) head of the sentence, with the verb itself as the head
of the VP. In this way, we can talk about a nite or non-nite sentence, one which is ultimately
headed by a nite or nonnite verb, respectively.
3
In addition to the complements of a head, a phrase may also contain modiers (or also called
adjuncts):
(11) a. Tom [
VP
[
VP
offered advice to his students] in his ofce].
b. Tom [
VP
[
VP
offered advice to his students] with love].
The PPs in his ofce or with love here provide further information about the action described
by the verb, but are not required as such by the verb. These phrases are optional and func-
tion as modiers, and they function to augment the minimal phrase projected from the head
verb offered. The VP which includes this kind of modier forms a maximal phrase. We might
say that the inner VP here forms a minimal VP which includes all the minimally required
complements, and the outer VP is the maximal VP which includes optional modiers.
What we have seen can be summarized as follows:
(12) a. Head: A lexical or phrasal element that is essential in forming a phrase.
b. Complement: A phrasal element that a head must combine with or a head select.
These include direct object, indirect object, predicative complement, and oblique
complement.
c. Modier: A phrasal element not selected by the verb functions as a modier to
the head phrase.
d. Minimal Phrase: A minimal phrase is the phrase including this head and all of
its complements.
3
See Chapter 5.2 for the detailed discussion of English verb form (VFORM) values including nite and nonnite.
51
e. Maximal Phrase: A maximal phrase is the phrase that includes complements as
well as modiers.
4.2 Differences between Complements and Modiers
Given these notions of complements and modiers, the question that follows is then how we can
distinguish between complements and modiers. There are several tests to determine whether a
phrase is a complement or a modier.
4
Obligatoriness: As hinted at already, complements are strictly-required phrases whereas mod-
iers are not. The examples in (13)(15) show that the verb placed requires an NP and a PP as
its complements, kept an NP and a PP or an AP, and stayed a PP.
(13) a. John placed Kim behind the garage.
b. John kept him behind the garage.
c. *John stayed Kim behind the garage.
(14) a. *John placed him busy.
b. John kept him busy.
c. *John stayed him busy.
(15) a. *John placed behind the counter.
b. *John kept behind the counter.
c. John stayed behind the counter.
In contrast, modiers are optional. Their presence is not required by the grammar:
(16) a. John deposited some money in the bank.
b. John deposited some money in the bank on Friday.
In (16b), the PP on Friday is optional here, serving as a modier.
Iterability: The possibility of iterating identical types of phrase can also distinguish between
complements and modiers. In general two or more instances of the same modier type can
occur with the same head, but this is impossible for complements.
(17) a. *The UN blamed global warming [on humans] [on natural causes].
b. Kim and Sandy met [in Seoul] [in the lobby of the Lotte Hotel] in March.
In (17a) on humans is a complement and thus the same type of PP on natural causes cannot
co-occur. Yet in Seoul is a modier and we can repeatedly have the same type of PP.
4
Most of the criteria we discuss here are adopted from Pollard and Sag (1987).
52
Do-so Test: Another reliable test often used to distinguish complements frommodiers is the do
so or do the same thing test. As shown in (18), we can use do the same thing to avoid repetition
of an identical VP expression:
(18) a. John deposited some money in the checking account and Mary did the same thing
(too).
b. John deposited some money in the checking account on Friday and Mary did the
same thing (too).
What we can observe in (18b) is that the VP did the same thing can replace either the minimal
phrase deposited some money in the checking account or the maximal phrase including the
modier on Friday. Notice that this VP can replace only the minimal phrase, leaving out the
modier.
(19) John deposited some money in the checking account on Friday and Mary did the
same thing on Monday.
From these observations, we can draw the conclusion that if something can be replaced by do
the same thing, then it is either a minimal or a maximal phrase. This in turn means that this
replacement VP cannot be understood to leave out any complement(s). This can be veried
with more data:
(20) a. *John [deposited some money in the checking account] and Mary did the same
thing in the savings account.
b. *John [gave a present to the student] and Mary did the same thing to the teacher.
Here the PPs in the checking account and to the student are both complements, and thus they
should be included in the do the same thing phrase. This gives us the following informal gener-
alization:
(21) Do-so Replacement Condition:
The phrase do so or do the same thing can replace a verb phrase which includes at
least any complements of the verb.
This condition explains why we cannot have another locative complement phrase in the
savings account or to the teacher in (20). The unacceptability of the examples in (22) also
supports this generalization about English grammar:
(22) a. *John locked Fido in the garage and Mary did so in the room.
b. *John ate a carrot and Mary did so a radish.
Constancy of semantic contribution: An adjunct can cooccur with a relatively broad range of
heads whereas a complement is typically limited in its distribution. Note the following contrast:
(23) a. Kim camps/jogs/mediates on the hill.
53
b. Kim jogs on the hill/under the hill/over the hill.
(24) a. Kim depends/relies on Sandy.
b. Kim depends on Sandy/*at Sandy/*for Sandy.
The semantic contribution of the adjunct on the hill in (23a) is independent of the head whereas
that of the complement on Sandy is idiosyncratically dependent upon the head.
Structural Difference: We could distinguish complements and modiers by tree structures,
too: complements combine with a lexical head (not a phrase) to form a minimal phrase whereas
modiers combine with a phrase to form a maximal phrase. This means that we have structures
of the following forms:
(25)
XP
XP
Modier
X Complement(s)
As represented in the tree structures, complements are sisters of the lexical head X, whereas
modiers are sisters of a phrasal head. This structural difference between complements and
modiers provides a clean explanation for the contrast in do-so test. Given that the verb ate
takes only an NP complement whereas put takes an NP and a PP complement, we will have the
difference in the two structures shown in (26):
(26) a.
VP
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
VP
.
.
.
.
.
.
.
PP
V NP
in the room
ate some food
b.
VP
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
V NP
PP
N
c. VP V (NP) (VP)
d. VP V NP AP
e. VP V NP NP
f. VP V S
g. AP A VP
h. PP P NP
i. VP Adv VP
One common property of all these rules is, as we have discussed, that every phrase has its own
head. In this sense, each phrase is the projection of a head, and thereby has the endocentricity.
However, we can ask the theoretical question of whether or not we can have rules like the
following, in which the phrase has no head at all:
(30) a. VP P NP
5
See (49) for the structural differences between with long hair and of linguistics.
6
The discussion of this section is based on Sag et al. (2003).
55
b. NP PP S
Nothing in the grammar makes such PS rules unusual, or different in any way from the set in
(29). Yet, if we allow such non-endocentric PS rules in which a phrase does not have a lexical
head, the grammar would then be too powerful to generate only the grammatical sentences of
the language.
Another limit that we can nd from the simple PS rules concerns an issue of redundancy.
Observe the following:
(31) a. *The problem disappeared the accusation.
b. The problem disappeared.
(32) a. *The defendant denied.
b. The defendant denied the accusation.
(33) a. *The boy gave the book.
b. The boy gave the baby the book.
What these examples show is that each verb has its own requirement for its complement(s).
For example, deny requires an NP, whereas disappear does not, and gave requires two NPs as
its complements. The different patterns of complementation are said to dene different subcate-
gories of the type verb. The specic pattern of complements is known as the subcategorization
requirement of each verb, which can be represented as following (IV: intransitive, TV: transitive,
DTV: ditransitive):
(34) a. disappear: IV,
b. deny: TV, NP
c. give: DTV, NP NP
In addition, in order to license the grammatical sentences in (31)(33), we need to have the
following three VP rules:
(35) a. VP IV
b. VP TV NP
c. VP DTV NP NP
We can see here that in each VP rule, only the appropriate verb can occur. That is, a DTV cannot
form a VP with the rules in (35a) or (35b): It forms a VP only according to the last PS rule.
Each VP rule thus also needs to specify the kind of verb that can serve as its head.
Taking these all together, we see that a grammar of the type just suggested must redundantly
encode the subcategorization information both in the lexical type of each verb (e.g., DTV) and
in the PS rule for that type of verb.
A similar issue of redundancy arises in accounting for subject-verb agreement:
(36) a. The bird devours the worm.
b. The birds devour the worm.
56
To capture the fact that the subject NP agrees with the predicate VP, we need to differentiate the
S rule into the following two:
(37) a. S NP
sing
VP
sing
(for (36)a)
b. S NP
pl
VP
pl
(for (36)b)
The two PS rules ensures that the singular (sing) subject combines with a singular VP whereas
the plural (pl) subject NP with a plural VP.
Descriptively, there is no problem with a grammar with many specic parts. From a theoret-
ical perspective, though, we have a concern about the the endocentricity and redundancy issues.
A more particular related question is that of how many PS rules English has. For example, how
many PS rules do we need to characterize English VPs?Presumably there are as many rules
as there are subcategories of verb.
We need to investigate the abstract content of PS rules, in order to develop a theoretical view
of them. For example, it seems to be the case that each PS rule must have a head. This will
disallow many possible PS rules which we can write using the rule format, from being actual
rules of any language.
In order to understand more about the structures that rules describe, we need two more no-
tions, intermediate category/phrase and specier (SPR). We motivate the idea of the inter-
mediate category, and then specier is a counterpart of it. Consider the examples in (38):
(38) a. Every photo of Max and sketch by his students appeared in the magazine.
b. No photo of Max and sketch by his students appeared in the magazine.
What are the structures of these two sentences? Do the phrases every photo of Max and sketch
by his students form NPs? It is not difcult to see sketch by his students is not a full NP by itself,
for if it was, it should be able to appear as subject by itself:
(39) *Sketch by his students appeared in the magazine.
In terms of the semantic units, we can assign the following structures to the above sentences, in
which every and no operate over the meaning of the rest of the phrase:
(40) a. [Every [[photo of Max] and [sketch by his students]]] appeared in the magazine.
b. [No [[photo of Max] and [sketch by his students]]] appeared in the magazine.
The expression photo of Max and sketch by his students are phrasal elements but not full NPs
so what are they? We call these intermediate phrases, notationally represented as N-bar or
N
. The phrase N
is thus intuitively bigger than a noun, but smaller than a full NP, in the sense
that it still requires a determiner from the class the, every, no, some, and the like.
The complementary notion that we introduce at this point is specier (SPR), which can
include the words just mentioned as well as phrases, as we illustrate in (41):
(41) a. [the enemys] [
N
destruction of the city]
b. [The enemy] [
VP
destroyed the city].
57
The phrase the enemys in (41a) and the subject the enemy in (41b) are semantically similar
in the sense that they complete the specication of the event denoted by the predicate. These
phrases are treated as the speciers of N
is a DP.
7
Now let us compare the syntactic structures of (41a) and (41b):
(43)
NP
DP
the enemys N PP
NP
VP
The enemy V
NP
but an NP.
58
phrase N
which in turn combines with the specier DP the enemys. In the S structure, the head
V combines with its complement the city and forms a VP. This resulting VP then combines with
the subject the enemy, which is also a specier. In a sense, the VP is an intermediate phrase that
requires a subject in order to be a full and complete S.
Given these similarities between NP and S structures, we can generalize over them as in (45),
where X is a variable over categories such as N, V, P, and other grammatical categories:
8
(45)
XP
Specier X
X Complement(s)
This structure in turn means the grammar now includes the following two rules:
9
(46) a. XP Specier, X
(Head-Specier Rule)
b. XP X, YP
(Head-Complement Rule)
These Head-Specier and Head-Complement Rules, which form the central part of X
-theory,
account for the core structure of NP as well as that of S. In fact, these two general rules can also
represent most of the PS rules we have seen so far. In addition to these two, we just need one
more rule:
10
(47) XP Modier, X
(Head-Modier Rule)
This Head-Modier Rule allows a modier to combine with its head as in the PS rule VP
VP Adv/PP, as presented in the following:
(48)
XP
Specier X
Modier
X Complement(s)
8
We can assume that the head of S is VP and that VP is an intermediate phrase in the sense that it still requires a
subject as its specier.
9
Unlike the PS rules we have seen so far, the rules here are further abstracted, indicated by the comma notation
between daughters on the right-hand side. We assume that the relative linear order of a head and complements etc. is
determined by a combination of general and language-specic ordering principles, while the hierarchical X
-structures
themselves are universal.
10
The comma indicates that the modier can appear either before the head or after the head as in always read books
or read books always.
59
One thing to notice in the Head-Complement Rule is that the head must be a lexical el-
ement. This in turn means that we cannot apply the Head-Modier Rule rst and then the
Head-Complement Rule. This explains the following contrast:
(49) a. the king [of Rock and Roll] [with a hat]
b. *the king [with a hat] [of Rock and Roll]
The badness of (49b) is due to the fact that the modier with a hat is combined with the head
king rst.
(50) a.
NP
.
.
.
.
.
.
.
.
.
.
.
.
DP N
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
the
N
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
PP
N
PP
with a hat
king of Rock and Roll
b.
NP
.
.
.
.
.
.
.
.
.
.
.
.
DP *N
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
the
N
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
PP
N
PP
but not an N or an NP. This will also account for the following contrast, too:
60
(52) A: Which student were you talking about?
B: The one with long hair.
B: *The one of linguistics with long hair.
The phrase of linguistics is the complement of student. This means the N-bar pronoun one
should include this. However, the modier with long hair cannot be within the N
.
There are several more welcome consequences that the three X
rules directly solve the endocentricity issue, for they refer to Head.
Assume that X is N, then we will have N, N
Specier X
[POS 1 ]
[POS 1 ]
Modier
X[POS 1 ] Complement(s)
The notation 1 shows that whatever value the feature has in one place in the structure, it has
the same value somewhere else. This is a representational tag, in which the number 1 has no
signicance: it could as easily be 7 or 437 . We provide more details of the formal feature system
in the following section.
So (53) indicates that the phrases POS value is identical to its head daughter, capturing
the headedness of each phrase: the grammar just does not allow any phrase without a head.
The redundancy issue mentioned above for agreement is now a matter of introducing another
feature, NUMBER. That is, with the new feature NUMBER, with values singular and plural,
we can add a detail to the Head-Specier Rule as following:
(54) XP Specier[NUMBER 1 ], X
[NUMBER 1 ]
The rule states that the subjects NUMBER value is identical with that of the predicate VPs
NUMBER value. The two rules in (37) are both represented in (54).
61
4.4 Lexicon and Feature Structures
In the previous section, we have seen that the properties of a lexical head determine the compo-
nents of the minimal phrase, in terms of complements, and that other properties of the head are
directly properties of the phrase. This information is encoded in a lexical entry, for each word
in the lexicon.
Every lexical entry at least includes phonological (but in practice, orthographic), morpho-
logical, syntactic, and semantic information. For example, the word puts will have at least the
following information:
(55) Minimal Lexical Information for puts:
a. phonological information: <puts>
b. syntactic information: verb, nite, 3rd singular
c. argument information: <agent
i
, theme
j
, location
k
>
d. semantic information: put
(i,j,k)
The phonological information is the information about how the word is pronounced; the syn-
tactic information indicates that this particular word is a verb and is in the 3rd singular present
(nite) form. The argument structure represents the number of arguments which the verb se-
lects, to indicate the participants that are minimally involved in the event expressed by the verb.
This argument information is linked to its more precise meaning as indicated by the indexes i,
j and k. These indexes refer to the participants denoted by the arguments. Finally, the semantic
structure represents that the verbs meaning relates three participants someone i who is doing
the action of putting, something j being put in a place, and someplace k it is put in. These lexi-
cal entries can be represented in a more systematic and precise way with the system of feature
structures, which we now introduce.
4.4.1 Feature Structures and Basic Operations
Most modern grammars rely on a representation of lexical information in terms of features
and their values.
11
We present here a formal and explicit way of representing it with feature
structures. Each feature structure is an attribute-value matrix (AVM):
(56)
_
_
Attribute1 value1
Attribute2 value2
Attribute3 value3
. . . . . .
_
_
The value of each attribute can be an atomic element, a list, a set, or a feature structure:
11
In particular, grammars such as Head-driven Phrase Structure Grammar (HPSG) and Lexical Functional Grammar
(LFG) are couched upon mathematically well-dened feature structure systems. The theory developed in this textbook
heavily relies upon the feature structure system of HPSG. See Sag et al. (2003).
62
(57)
_
_
Attribute1 atomic
Attribute2
Attribute3
_ _
Attribute4
_
. . .
_
_
_
One important property of every feature structure is that it is typed.
12
That is, each feature
structure is relevant only for a given type. A simple illustration can show why feature structure
needs to be typed. The upper left declaration in italics is the type of the feature structure:
(58) a.
_
_
university
NAME kyunghee univ.
LOCATION seoul
_
_
b.
*
_
_
university
NAME kyunghee univ.
MAJOR linguistics
_
_
The type university may have many properties, including its name and location, but having a
MAJOR, as a subject of study, is inappropriate. In the linguistic realm, we might declare that
TENSE is appropriate only for verb, for example.
Now consider the following example of a typed feature structure, information about one
author of this book:
(59)
_
_
author
NAME kim
SONS Edward, Richard
HOBBIES
_
swimming, jogging, reading, . . .
_
ADVANCED-DEGREE
_
_
FIELD linguistics
AREA syntax-semantics
YEAR 1996
_
_
_
_
This illustrates the different types of values that attributes (feature names) may have. Here, the
value of the attribute NAME is atomic, whereas the value of SONS is a list which represents
something relative about the two values, in this case that one is older than the other. So, for
example youngest son would be the right-most element in the list value of SONS. Meanwhile,
the value of HOBBIES is a set, showing that there is no signicance in the relative ordering.
Finally, the value of the feature ADVANCED-DEGREE is a feature structure which in turn has
three attributes.
One useful notion in the feature structure is structure-sharing, which we have already seen
above in terms of the 1 notation (see (53)). This is to represent cases where two features (or
12
Even though every feature structure is typed in the present grammar, we will not specify the type of each feature
structure unless it is necessary for the discussion.
63
attributes) have an identical value:
(60)
_
_
individual
NAME kim
TEL 1
SONS
_
_
_
individual
NAME richard
TEL 1
_
_
,
_
_
individual
NAME edward
TEL 1
_
_
_
_
_
For the type individual, attributes such as NAME and TEL and SONS are appropriate. (60)
represents a situation in which the particular individual kimhas two sons, and their TEL attribute
has a value which is the same as the value of his TEL attribute, whatever the value actually is.
In addition to this, the notion of subsumption is also important in the theoretical use of
feature structures; the symbol represents subsumption. The subsumption relation concerns
the relationship between a feature structure with general information and one with more specic
information. In such a case, the general one subsumes the specic one. Put differently, a feature
structure A subsumes another feature structure B if A is not more informative than B.
(61)
A:
_
individual
NAME kim
_
B:
_
_
individual
NAME kim
TEL 961-0892
_
_
In (62), A represents more general information than B. This kind of subsumption relation is
used to represent partial information, for in fact we cannot represent the total information
describing all possible worlds or states of affairs. In describing a given phenomenon, it will be
more than enough just to represent the particular or general aspects of the facts concerned. Each
small component of feature structure will provide partial information, and as the structure is
built up, the different pieces of information are put together.
The most crucial operation in feature structures is unication, represented by
. Feature
unication means that two compatible feature structures are unied, conveying more coherent
and rich information. Consider the feature structures in (62); the rst two may unify to give the
third:
(62)
_
individual
NAME kim
_
_
individual
TEL 961-0892
_
_
_
individual
NAME kim
TEL 961-0892
_
_
The two feature structures are unied, resulting in a feature structure with both NAME and TEL
information. However, if two feature structures have incompatible feature values, they cannot
be unied:
64
(63)
_
individual
NAME edward
_
_
individual
NAME richard
_
*
_
_
individual
NAME edward
NAME richard
_
_
Since the two smaller feature structures have different NAME values, they cannot be unied.
Unication will make sure that information is consistent as it is built up in the analysis of a
phrase or sentence.
4.4.2 Feature Structures for Linguistic Entities
Any individual or entity including a linguistic expression can be represented by a feature struc-
ture. For example, the word puts, whose general type is verb, can have a feature structure like
the following:
13
(64)
_
_
verb
PHON puts
SYN
_
_
POS verb
VFORM n
_
_
ARG-ST
_
[agt]
i
, [th]
j
, [loc]
k
_
SEM
_
_
PRED put-relation
AGENT i
THEME j
LOCATION k
_
_
_
_
This feature structure has roughly the same information as the informal representation in (55).
The verb puts, like any verb, has its own phonological (PHON) value, syntactic (SYN), ar-
gument structure (ARG-ST), and semantic (SEM) information. The SYN attribute indicates
that the POS (part of speech) value is verb, that it has a nite verbal inectional form value
(VFORM).The ARG-ST attribute indicates that the verb selects for three arguments (with the-
matic roles agent, theme, location), which will be realized as the subject and two complements
in the full analysis. The SEM feature represents the information this verb denotes the predicate
relation put-relation, whose three participants are linked to the elements in the ARG-ST via the
indexing values i, j, and k.
One thing to note here is that since there are some cases where we have difculties in assign-
ing a specic named semantic role to a selected argument discussed in Chapter 3, we typically
just indicate the number of arguments each predicate is selecting in ARG-ST: we underspecify
13
Later on, we will not represent the PHON and SEM values unless relevant to the discussion at hand.
65
the information unless it is necessary to show more details. So, for example, verbs like smile,
devour and give will have the following ARG-ST representations, respectively:
(65) a.
_
ARG-ST [ ]
_
b.
_
ARG-ST [ ], [ ]
_
c.
_
ARG-ST [ ], [ ], [ ]
_
One-place predicates like smile select for just one argument, two-place predicates like devour
take two arguments, and three-place predicates take three arguments. Eventually, the arguments
selected by each predicate are linked to grammatical functions, to the core semantic properties,
and to other parts of the representation of the grammatical properties.
4.4.3 Argument Realization
Each element on the ARG-ST list is realized as SPR (specier) or COMPS (complements),
through one of the rules in (46).
14
In general, the basic pattern is that the rst element on the list
is realized as subject and the rest as complements:
(66) Argument Realization Constraint (ARC, rst approximation):
The rst element on the ARG-ST list is realized as SPR, the rest as COMPS in
syntax.
This realization is obligatory in English; for example, the three arguments of put are realized as
subject and complements, with the putter (agent) as subject:
15
(67) a. John put the book in the box.
b. *John put in the box.
c. *In the box put John the book.
d. #The book put John in the box.
We see that the arguments selected by a lexical head should be all realized as SPR and COMPS,
which are combined in the notion of valence (VAL) features.
16
More formally, we can represent
this constraint as applied to put as the following feature structure:
(68)
_
_
VAL
_
_
SPR 1 NP
COMPS 2 NP, 3 PP
_
_
ARG-ST 1 , 2 , 3
_
_
The boxed tags show the different identities in the overall structure. For example, the rst el-
ement of ARG-ST and of SPR have the boxed tag 1 , ensuring that the two are identical. The
14
Once again, remember that the term SPR includes subject as well as the nouns specier.
15
The notation # indicates that the structure is technically well-formed from a syntactic perspective, but semantically
anomalous.
16
The term valence refers to the number of arguments that a lexical item can combine with, to make a syntactically
well-formed sentence, often along with a description of the categories of those constituents. It is inspired by the notion
of valence as used in atomic theory in chemistry.
66
general ARC constraint blocks examples like (67c) in which the location argument is realized
as the subject, as shown in (69):
(69) *
_
_
VAL
_
_
SPR 3 PP
COMPS 1 NP, 2 NP
_
_
ARG-ST 1 , 2 , 3
_
_
This violates the ARC, which requires the rst element of ARG-ST be realized as the SPR (the
subject of a verb or the specier of a noun).
Notice that the arguments can be realized into different categories, depending on the proper-
ties of the given verb:
(70) a. The election results surprised everybody.
b. That he won the election surprised everybody.
The data indicate that verbs like surprise will take two arguments, but the rst argument can be
realized either as an NP subject as in (70a) or a CP subject as in (70b). This difference in the
argument realization can be represented as the following, respectively:
(71) a.
_
_
VAL
_
_
SPR 1 NP
COMPS 2 NP
_
_
ARG-ST 1 , 2
_
_
b.
_
_
VAL
_
_
SPR 1 CP
COMPS 2 NP
_
_
ARG-ST 1 , 2
_
_
Though there is no difference in terms of the number of arguments that surprise select, the
arguments can be realized in a different phrase. As the book goes on, we will see how the
argument realization is further constrained by the lexical properties of the verb in question or
by other grammatical components.
4.4.4 Verb Types and Argument Structure
As mentioned earlier, lexical elements in the classes V, A, N, and P, select one or more comple-
ment(s) to form a minimal phrase. With the construct of ARG-ST, we know that every lexical
element has ARG-ST information which will be realized in surface form through the SPR and
COMPS values. Verb types can be differentiated by looking only at the COMPS value since
every verb will have one SPR (subject) element. This is exactly the way that verbs are differen-
tiated using the traditional notion of subcategorization.
Intransitive: This is a type of verb that does not have any COMPS:
(72) a. John disappeared.
b. *John disappeared Bill.
67
(73) a. John sneezed.
b. *John sneezed the money.
These verbs have no COMPS elementthe list is necessarily empty. Such a verb will have just
one argument that is realized as subject:
(74)
_
_
disappear
VAL
_
_
SPR 1 NP
COMPS
_
_
ARG-ST 1
_
_
Linking verbs: Verbs such as look, seem, remain, and feel require different complements that
are typically of category AP:
(75) a. The president looked [weary].
b. The teacher became [tired of the students].
c. The lasagna tasted [scrumptious].
d. John remained [somewhat calm].
e. The jury seemed [ready to leave].
These verbs also can select other phrases (here, NP):
(76) a. John became a success.
b. John seemed a fool.
c. John remained a student.
Though each verb may select different types of phrases, they all at least select a predicative
complement, where a property is ascribed to the subject. (Compare John remained a student
and John revived a student.) This subcategorization requirement can be represented as follows:
(77)
_
_
become
VAL
_
_
SPR 1 NP
COMPS 2 AP/NP[PRD +]
_
_
ARG-ST 1 , 2
_
_
This kind of verb selects two arguments: one is canonically an NP to be realized as the subject
and the other is any phrase (XP) that can function as a predicate (PRD +) (see also the examples
in (84)). Of course, this presupposes an accurate characterization of which phrases can be [PRD
+], which we simply assume here.
Transitive verbs: Unlike linking verbs, pure transitive verbs select a referential, non-predicative
NP as their complement, functioning as direct object:
(78) a. John saw Fred.
b. Alice typed the letter.
68
c. Clinton supported the health care bill.
d. Raccoons destroyed the garden.
Such verbs will have the following lexical information:
(79)
_
_
destroy
VAL
_
_
SPR 1 NP
COMPS 2 NP
_
_
ARG-ST 1 , 2
_
_
Ditransitive: There are also ditransitive verbs that require IO and DO:
(80) a. The school board leader asked the students a question.
b. The parents bought the children non-ction novels.
c. John taught new students English Syntax.
Such verbs have three arguments: one subject and two complement NPs functioning as indi-
rect and direct object, respectively:
(81)
_
_
teach
VAL
_
_
SPR 1 NP
COMPS 2 NP, 3 NP
_
_
ARG-ST 1 , 2 [goal], 3 [theme]
_
_
These verbs typically have a related realization with the goal argument realized as a dative
PP complement:
(82) a. The school board leader asked a question of the students.
b. The parents bought non-ction novels for the children.
c. John taught English Syntax to new students.
In this realization, the second argument has the theme role while the third one has the goal role.
(83)
_
_
teach
VAL
_
_
SPR 1 NP
COMPS 2 NP, 3 PP
_
_
ARG-ST 1 , 2 [theme], 3 [goal]
_
_
Complex Transitive: There is another type of transitive verb which selects two complements,
one functioning as a direct object and the other as a predicative phrase (NP, AP, or VP), describ-
ing the object:
(84) a. John regards Bill as a good friend.
b. The sexual revolution makes some people uncomfortable.
c. Ad agencies call young people Generation X-ers.
69
d. Historians believe FDR to be our most effective president.
In (84a), the predicative PP as a good friend follows the object Bill; in (84b), the AP uncom-
fortable serves as a predicate phrase of the preceding object some people. In (84c), the NP
Generation X-ers is the predicative phrase. In (84d), the predicative phrase is an innitive VP.
Just like linking verbs, these verbs require a predicative ([PRD +]) XP as complement:
(85)
_
_
call
VAL
_
_
SPR 1 NP
COMPS 2 NP, 3 XP
_
_
ARG-ST 1 , 2 , 3 [PRD +]
_
_
This means that the verbs in (84) all select an object NP and an XP phrase that function as a
predicate.
Even though these ve types of verb that we have seen so far represent many English verb
types, there are other verbs that do not t into these classes; for instance, the use of the verb
carry in (86).
(86) a. *John carried to the door.
b. John carried her on his back.
The examples in (86) illustrate that carried requires an NP and a PP, as represented in the feature
structure:
(87)
_
_
carry
VAL
_
_
SPR 1 NP
COMPS 2 NP, 3 PP
_
_
ARG-ST 1 [agt], 2 [th], 3 [loc]
_
_
The PP here cannot be said to be predicate of the object her; it denotes the location to which
John carries her.
Of course, there are various other verb types that we have not described here, in terms of
complementation patterns. As the book goes on, we will see yet more different types of verbs.
70
4.5 Exercises
1. For each of the verbs below, use it in a grammatical example; write down the example
and provide its lexical entry including the COMPS value.
(i) appear, consider, award, borrow, owe, explain, introduce, discuss, tell, say
2. Provide tree structures for the following pairs of sentences while checking the gram-
matical function of each phrase with valid distributional tests. In doing so, state what
differences we need to represent for the underlined expressions.
(i) a. Tom locked Fido in the garage.
b. Tom bathed Fido in the garage.
(ii) a. Tom placed it under the table.
b. Tom played it under the table.
(iii) a. I wonder if you will come back tomorrow.
b. You would have a reply if you come back tomorrow.
3. For each example below, draw its structure and then give the lexical entry (SPR and
COMPS value) of the main verb.
(i) a. Tom hid the manuscript in the cupboard.
b. Fred hired Sharon to change the oil.
c. They pushed the prisoners into the truck.
d. Frank hopes to persuade Harry to make the cook wash the dishes.
e. George mailed the attorney his photograph of the accident.
f. Tom keeps asking Karens sister to buy the car.
g. Jane left the book on the table.
h. We have not conrmed whether the ight had been booked.
i. We saw him beaten by the champion.
j. They conned his remarks to the matter under discussion.
4. The verbs in the following examples are used incorrectly. Correct the errors or replace
the verb with another one, and write out each new example. In addition, provide the
COMPS value for each verb (in its use in your grammatical examples).
(i) a. *Oliver ascribed his longevity there.
b. *Oliver mentioned Charles the problem.
c. *Oliver ned ten pounds to the prisoner.
d. *Oliver drove me a lunatic.
e. *Oliver addressed the king the letter.
f. *Oliver absented his brother from the meeting.
5. Draw tree structures for the following two sentences. In particular, provide detailed NP
structures using the intermediate phrase N
:
(i) a. The students of English from Seoul faced many issues in the process of
interpreting, transcribing, and editing the poems.
71
b. The love of my life and father of my children would never do such a thing.
c. The museum displayed no painting by Miro or drawing by Klee.
d. By law, every dog and cat in the area has to be spayed or neutered.
6. Read the following texts and provide the ARG-ST, SPR and COMPS values of the un-
derlined expressions.
(i) Learning to use a language freely and fully is a lengthy and effortful pro-
cess. Teachers cannot learn the language for their students. They can set
their students on the road, helping them to develop condence in their own
learning powers. Then they must wait on the sidelines, ready to encourage
and assist.
(ii) Deep ecologists put a reign on human exploitation of natural resources
except to satisfy vital needs. Thus, the use of a eld by an African tribe to
growgrain for survival is an example of a vital need whereas the conversion
of a swamp to an exclusive golf course would not. Rest assured that much
of the mining, harvesting, and development of our technological age would
not meet the requirement of this principle. Rather than being concerned
about how to raise automobile production, this ethic would be interested
in solving the problem of human mobility in a way that would not require
the disruption of highways, roads, and parking lots. It rebels against an
industrialist world view: Before it is possessed and used, every plant is a
weed and every mineral is just another rock.
17
17
From http://www.unitedearth.com.au/deepecology.html
72
5
More on Subjects and Complements
5.1 Grammar Rules and Principles
As we have seen in the previous chapter, the arguments in ARG-ST are realized as the syntactic
elements SPR (subject of a verb and determiner of a noun) and COMPS. The X
rules control
their combination with a relevant head:
(1) a. XP Specier, Head
b. XP Head, Complement(s)
c. XP Modier, Head
The rule (1a) represents the case where a head combines with its specier (e.g., a VP with
its subject and an N
with its determiner), whereas (1b) says that a head combines with its
complement(s) and forms a phrase. (1c) allows a combination of a head with its modier. As
noted earlier, in order to guarantee that the heads POS (part of speech) value is identical with
its mother phrase, we need to introduce the category variable X and the feature POS:
(2) a. Head-Specier Rule:
XP
_
POS 1
_
Specier, X
[POS 1 ]
b. Head-Complement Rule:
XP
_
POS 1
_
X[POS 1 ], Complement(s)
c. Head-Modier Rule:
XP
_
POS 1
_
Modier, XP
_
POS 1
_
The POS feature is thus a head feature which passes up to a mother phrase from its head
daughter, as shown in (3):
(3)
VP[POS verb]
V[POS verb]
PP
73
This percolation from a head to its mother is ensured by the following Head Feature Principle:
(4) The Head Feature Principle (HFP):
A phrases head feature (e.g., POS, VFORM, etc.) is identical with that of its head.
The HFP thus ensures that every phrase has its own lexical head with the identical POS value.
The HFP will apply to any features that we declare to be head features, VFORMbeing another.
The grammar does not allow hypothetical phrases like the following:
(5)
*VP[POS verb]
.
.
.
.
.
.
.
.
.
.
.
A[POS adj] PP
We have not yet spelled out clearly what ensures a lexical head to combine not just with one
of its complements but with all of its COMPS elements. Consider the following examples:
(6) a. Kim put the book in the box.
b. *Kim put the book.
c. *Kim put in the box.
As seen from the contrast here and as noted in the lexical entry in (7), the verb put selects two
complements and must combine with all of its complements.
(7)
_
_
HEAD| POS verb
VAL
_
_
SPR NP
COMPS NP, PP
_
_
_
_
We can also see that a nite verb must combine with its subject:
(8) a. *Is putting the book in the box.
b. *Talked with Bill about the exam.
Such combinatorial requirements can be formally stated in the revised grammar rules as given
in (9):
(9) a. Head-Specier Rule:
XP
_
SPR
_
1 , H[SPR 1 ]
b. Head-Complement Rule:
XP
_
COMPS
_
H[COMPS 1 ,. . . , n ], 1 ,. . . , n
c. Head-Modier Rule:
XP [MOD 1 ], 1 H
The grammar rules here are well-formedness conditions on possible phrases in English, indi-
cating what each head combines with and then what happens as the result of the combination.
74
For example, in (9a) when a head, requiring a SPR, combines with it, we have a well-formed
head-specier phrase with the SPR value discharged; in (9b), a head combines with all of its
COMPS value, it forms a Head-Complement phrase with no further COMPS value; in (9c),
when a modier combines (carrying the MOD feature) with the head it modies, the resulting
phrase forms a well-formed head-modier phrase.
1
These three grammar rules, interacting with the general principles such as the HFP, license
grammatical sentences in English. Let us consider one example in a little more detail:
2
(10)
S
_
_
HEAD 4 | POS verb
SPR
COMPS
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
Subj
H
1 NP
VP
_
_
HEAD 4 | POS verb
SPR 1
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
H
C
C
Kim
V
_
_
HEAD 4 | POS verb
SPR 1
COMPS 2 , 3
_
_
2 NP
3 PP
nonn
NP
VP[VFORM n]
One thing we need to remember is that the two participle forms (present and past) have many
different uses, in different constructions, as partially exemplied in (21) and (22).
(21) Usages of the present participle (ing) form:
a. He is writing another long book about beavers.
(part of the the progressive aspect construction)
b. Broadly speaking, the project was successful.
(used as a sentence modier)
c. He is proud of his sons passing the bar exam.
(used in a gerundive construction)
(22) Usages of the past participle (en) form:
a. The chicken has eaten.
(part of the perfect aspect construction)
b. The chicken was eaten.
(part of the passive voice construction)
c. Seen from this perspective, there is no easy solution.
(used as a sentence modier)
Some of these usages have been introduced as VFORM values for the ease of exposition (cf.
Gazdar et al. (1985) or Ginzburg and Sag (2000).) though strictly speaking, there are only two
participle forms in English, which each have several functions or constructional usages.
Every verb will be specied for a value of the head feature VFORM. For example, let us
consider a simple example like The student knows the answer. Here the verb knows will have
the following lexical information:
(23)
_
_
knows
HEAD
_
_
POS verb
VFORM pres
_
_
VAL
_
_
SPR
_
1 NP
_
COMPS
_
2 NP
_
_
_
ARG-ST
_
1 , 2
_
_
_
78
This [VFORM pres] value will be projected to the S in accordance with the HFP, as represented
in the following:
(24)
S
_
VFORM pres
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
NP
VP
_
VFORM pres
_
.
.
.
.
.
.
.
.
.
The student
V
_
VFORM pres
_
NP
_
keep
HEAD| POS verb
COMPS
_
VP[ing]
_
_
_
b.
_
_
make
HEAD| POS verb
COMPS
_
NP, VP[bse]
_
_
_
Such lexical specications on the VPs VFORM value will make sure that these verbs only
combine with a VP complement with the appropriate VFORM value. The following structure
represents one example:
79
(27)
S
_
_
HEAD 4
_
_
POS verb
VFORM pst
_
_
SPR
COMPS
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
1 NP
VP
_
_
HEAD 4
SPR 1 NP
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
N
V
_
_
HEAD 4
SPR 1 NP
COMPS 2 VP
_
_
2 VP
_
_
HEAD 5 | VFORM ing
SPR 2 NP
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
John kept
V
_
_
HEAD 5
SPR 2 NP
COMPS 5
_
_
5 NP
_
eager
HEAD| POS adj
SPR NP
COMPS VP[VFORM inf ]
_
_
b.
_
_
fond
HEAD| POS adj
SPR NP
COMPS PP[PFORM of ]
_
_
Such lexical entries will project sentences like the following:
81
(31)
S
_
VFORM pres
_
_
_
_
_
_
_
_
_
_
_
_
_
3 NP
VP
_
VFORM pres
_
.
.
.
.
.
.
.
.
.
.
.
.
Monkeys
V
_
_
VFORM pres
COMPS 2
_
_
2 AP
are
A
_
COMPS 1
_
1 VP
_
VFORM inf
_
.
.
.
.
.
.
.
.
.
.
.
.
eager
V
_
VFORM inf
_
VP
_
proximity
HEAD| POS noun
SPR
_
1 DP
_
COMPS
_
2 PP[PFORM to]
_
_
_
b.
_
_
faith
HEAD| POS noun
SPR
_
1 DP
_
COMPS
_
2 PP[PFORM in]
_
_
_
Though many more details remain to be covered for the various complement types of lexical cat-
egories, the discussion so far has given an idea of what kinds of complement lexical categories
select for.
5.3 Feature Specications for the Subject
In general, most verbs select a regular NP as subject:
(34) a. John/Some books/The spy disappeared.
b. The teacher/The monkey/He fooled the students.
However, as noted in the previous chapter, certain English verbs select only it or there as subject:
(35) a. It/*John/*There rains.
b. There/*The spy lies a man in the park.
The pronouns it and there are often called expletives, indicating that they do not have or con-
tribute any meaning. The use of these expletives is restricted to particular contexts or verbs,
though both forms have regular pronoun uses as well. One way to specify such lexical speci-
cations for subjects is to make use of a form value specication for nouns: all regular nouns
[NFORM norm(al)] as default specication; overall we classify nouns as having three different
NFORM values: normal, it, and there. Given the NFORM feature, we can have the following
lexical entries for the verbs above:
(36) a.
_
_
rained
SPR NP[NFORM it]
COMPS
_
_
4
DP covers not only simple determiners like a, the, and that, but also includes a possessive phrase like Johns. In
Chapter 6 we cover NP structures in detail.
83
b.
_
_
fooled
SPR NP[NFORM norm
COMPS NP
_
_
We can also observe that only a limited set of verbs require their subject to be [NFORM
there]:
5
(37) a. There exists only one truly amphibian mammal.
b. There arose a great storm.
(38) a. There exist few solutions which are cost-effective.
b. There is a riot in the park.
c. There remained just a few problems to be solved.
The majority of verbs do not allow there as subject:
(39) a. *There runs a man in the park.
b. *There sings a man loudly.
For the sentences with there subjects, we rst consider the verb forms which have regular sub-
jects. A verb like exist in (37a) takes one argument when used in an example like Only one truly
amphibian mammal exists, and the argument will be realized as the SPR, as dictated by the
entry in (40a). In addition, such verbs can introduce there as the subject, through the Argument
Realization option given in (40b), which is the form that occurs in the structure of (37a).
(40) a.
_
_
exists
SPR 1 NP
COMPS
ARG-ST 1 NP
_
_
b.
_
_
exists
SPR 1 NP[NFORM there]
COMPS 2 NP
ARG-ST 1 NP, 2 NP
_
_
5.4 Clausal Complement or Subject
5.4.1 Verbs Selecting a Clausal Complement
We have seen that the COMPS list includes predominantly phrasal elements. However, there are
verbs selecting not just a phrase but a whole clause as a complement, either nite or nonnite.
For example, consider the complements of think or believe:
(41) a. I think (that) the press has a check-and-balance function.
b. They believe (that) Charles Darwins theory of evolution is just a scientic theory.
5
Some verbs such as arise or remain sound a little archaic in these constructions.
84
The C (complementizer) that here is optional, implying that this kind of verb selects for a nite
complement clause of some type, which we will notate as a [VFORM n] clause. That is, these
verbs will have one of the following two COMPS values:
(42) a.
_
COMPS
_
S[VFORM n]
_
_
b.
_
COMPS
_
CP[VFORM n]
_
_
If the COMPS value only species a VFORM value, the complement can be either S or CP. This
means that we can subsume these two uses into the following single lexical entry, suppressing
the category information of the sentential complement:
6
(43)
_
_
believe
HEAD| POS verb
COMPS
_
[VFORM n]
_
_
_
We can also nd somewhat similar verbs like demand and require:
(44) a. John demanded [that she stop phoning him].
b. The rules require [that the executives be polite].
Unlike think or believe, these verbs which introduce a subjunctive clause typically only take a
CP[VFORM bse] as their complement: the nite verb itself is actually in the bse form. Observe
the structure of (44b):
6
Although the categories V or VP are also potentially specied as [VFORM n], such words or phrases cannot
be complements of verbs like think or believe. This is because complements are typically saturated phrases, with no
unsatised requirements for complements or speciers. While S and CP are saturated categories projected from V, VP
and V are not saturated.
85
(45)
S
NP
VP
_
COMPS
_
.
.
.
.
.
.
.
.
.
.
.
.
The rules
V
_
COMPS 2
_
2 CP
_
VFORM bse
_
.
.
.
.
.
.
.
.
.
.
.
.
require
C
_
VFORM bse
_
S
_
VFORM bse
_
.
.
.
.
.
.
.
.
.
.
.
.
_
that
HEAD
_
_
POS comp
VFORM 1
_
_
SPR
COMPS S[VFORM 1 ]
_
_
This lexical information will then allow us to pass on the VFORM value of S to the head C
and then be percolated up to the CP according to the HFP. This encodes the intuition that a
complementizer agrees with regard to VFORM value with its complement sentence.
There are also verbs which select a sequence of an NP followed by a CP as complements. NP
and CP are abbreviations for feature structure descriptions that include the information [POS
noun] and [POS comp], respectively:
(47) a. Joe warned the class that the exam would be difcult.
b. We told Tom that he should consult an accountant.
c. Mary convinced me that the argument was sound.
The COMPS value of such verbs will be as in (48):
86
(48)
_
COMPS NP, CP[VFORM n]
_
In addition to the that-type of CP, there is an innitive type of CP, headed by the comple-
mentizer for. Some verbs select this nonnite CP as the complement:
(49) a. Tom intends for Sam to review that book.
b. John would prefer for the children to nish the oatmeal.
The data show that verbs like intend and prefer select an innitival CP clause. The structure of
(49a) is familiar, but now has a nonnite VFORM value within it:
(50)
S
NP
VP
_
COMPS
_
.
.
.
.
.
.
.
.
.
.
.
Tom
V
_
COMPS 2
_
2 CP
_
VFORM inf
_
.
.
.
.
.
.
.
.
.
.
.
intends
C
_
VFORM inf
_
S
_
VFORM inf
_
for NP
VP
_
VFORM inf
_
.
.
.
.
.
.
.
.
.
.
.
Sam
V
_
VFORM inf
_
VP
_
intends
HEAD
_
_
POS verb
VFORM pres
_
_
COMPS CP[VFORM inf ]
_
_
87
To ll out the analysis, we need explicit lexical entries for the complementizer for and for to,
which we treat as an (innitive) auxiliary verb. In fact, to has a distribution very similar to the
nite modal auxiliaries such as will or must, differing only in the VFORM value (see section
8.3.4).
(52) a.
_
_
for
HEAD
_
_
POS comp
VFORM inf
_
_
COMPS S[VFORM inf ]
_
_
b.
_
_
to
HEAD
_
_
POS verb
VFORM inf
_
_
COMPS VP[VFORM bse]
_
_
Just like the complementizer that, the complementizer for selects an innitival S as its comple-
ment, inheriting its VFORM value too. The evidence that the complementizer for requires an
innitival S can be found from coordination data:
(53) a. For John to either [make up such a story] or [repeat it] is outrageous.
(coordination of bse VPs)
b. For John either [to make up such a story] or [to repeat it] is outrageous.
(coordination of inf VPs)
c. For [John to tell Bill such a lie] and [Bill to believe it] is outrageous.
(coordination of inf Ss)
Given that only like categories (constituents with the same label) can be coordinated, we can
see that base VPs, innitival VPs, and innitival Ss are all constituents.
7
One thing to note here is that the verbs which select a CP[VFORM inf ] complement can also
take a VP[VFORM inf ] complement:
(54) a. John intends to review the book.
b. John would prefer to nish the oatmeal.
By underspecifying the category information of complements, we can generalize this subcate-
gorization information:
(55)
_
_
intend
HEAD| POS verb
COMPS
_
[VFORM inf ]
_
_
_
7
Tensed VPs can be coordinated regardless of their different tense values as in Kim [alienated cats] and [beats his
dog].
88
Since the specication [VFORM inf ] is quite general, it can be realized either as CP[VFORM
inf ] or VP[VFORM inf ].
However, this does not mean that all verbs behave alike: not all verbs can take variable
complement types such as an innitival VP or S. For examples, try, tend, hope, and others
select only a VP[inf ] as attested by the data:
(56) a. Tom tried to ask a question.
b. *Tom tried for Bill to ask a question.
(57) a. Tom tends to avoid confrontations.
b. *Tom tends for Mary to avoid confrontations.
(58) a. Joe hoped to nd a solution.
b. *Joe hoped for Beth to nd a solution.
Such subcategorization differences are hard to predict just from the meaning of verbs: they are
simple lexical specications which language users need to learn.
There is another generalization that we need to consider with respect to the property of verbs
that select a CP: most verbs that select a CP can at rst glance select an NP, too:
(59) a. John believed it/that he is honest.
b. John mentioned the issue to me/mentioned to me that the question is an issue.
Should we have two lexical entries for such verbs or can we have a simple way of representing
such a pattern? To reect such lexical patterns, we can assume that English parts of speech come
in families and can protably be analyzed in terms of a type hierarchy as following:
8
(60)
part-of-speech
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
nominal
verbal
adj prep . . .
noun comp verb
According to the hierarchy, the type nominal is a supertype of both noun and comp. In accor-
dance with the basic properties of systems of typed feature structures, an element specied as
[POS nominal] can be realized either as [POS noun] or [POS comp]. These will correspond to
the phrasal types NP and CP, respectively.
9
The hierarchy implies that the subcategorization pattern of English verbs will refer to (at
least) each of these types. Consider the following patterns:
(61) a. She pinched [his arm] as hard as she could.
8
This type hierarchy is adopted from Kim and Sag (2006).
9
For the function of verbal, see Chapter 12.
89
b. *She pinched [that he feels pain].
(62) a. We hope [that such a vaccine could be available in ten years].
b. *We hope [the availability of such a vaccine in ten years].
(63) a. Cohen proved [the independence of the continuum hypothesis].
b. Cohen proved [that the continuum hypothesis was independent].
The part-of-speech type hierarchy in (60) allows us to formulate simple lexical constraints that
reect these subcategorization patterns, making reference to noun, verbal, and nominal:
(64) a.
_
ARG-ST NP, NP[POS noun], . . .
_
b.
_
ARG-ST NP, CP[POS comp], . . .
_
c.
_
ARG-ST NP, [POS nominal], . . .
_
In each class, the ARG-ST list species the argument elements that the verbs select (in the order
Subject, Direct Object, . . . ). The POS value of a given element is the part-of-speech type that
a word passes on to the phrases it projects. These three patterns illustrate that English transitive
verbs come in at least three varieties.
5.4.2 Verbs Selecting a Clausal Subject
In addition to CP as a complement, we also nd some cases where a CP is the subject of a verb:
(65) a. [John] bothers me.
b. [That John snores] bothers me.
(66) a. [John] loves Bill
b. *[That John snores] loves Bill.
The contrast here means that verbs like bother can have two realizations of the ARG-ST,
whereas those like love allow only one:
(67) a.
_
_
bothers
SPR 1 [POS nominal]
COMPS 2 NP
ARG-ST 1 , 2
_
_
b.
_
_
loves
SPR 1 NP
COMPS 2 NP
ARG-ST 1 , 2
_
_
These different realizations all hinge on the lexical properties of the given verb, and only some
verbs allow the dual realization described by (67a).
A clausal subject is not limited to a nite that-headed CP, but there are other clausal types:
90
(68) a. [That John sold the ostrich] surprised Bill.
(that-clause CP subject)
b. [(For John) to train his horse] would be desirable.
(innitival CP or VP subject)
c. [That the king or queen be present] is a requirement on all Royal weddings.
(subjunctive that-clause CP subject)
d. [Which otter you should adopt rst] is unclear.
(wh-question CP subject)
Naturally, each particular predicate dictates which kinds of subjects are possible, as in (68),
and which are not, as in (69):
(69) a. *That Fred was unpopular nominated Bill.
b. *That Tom missed the lecture was enjoyable.
c. *For John to remove the mother is undeniable.
d. *How much money Gordon spent is true.
For example, the difference between the two verbs nominate and surprise can be seen in these
partial lexical entries:
(70) a.
_
_
nominate
VAL
_
_
SPR 1 NP
COMPS 2 NP
_
_
ARG-ST 1 , 2
_
_
b.
_
_
surprise
VAL
_
_
SPR 1 [POS nominal]
COMPS 2 NP
_
_
ARG-ST 1 , 2
_
_
Unlike nominate, the rst argument of surprised can be a nominal. This means that its subject
can be either an NP or a CP.
5.4.3 Adjectives Selecting a Clausal Complement
Like verbs, certain adjectives can also select CPs as their complements. For example, condent
and insistent select a nite CP, whereas eager selects an innitival CP:
(71) a. Tom is condent [that the elephants respect him].
b. Tom is insistent [that the witnesses be truthful].
(72) a. Tom seems eager [for her brother to catch a cold].
b. Tom seems eager [to catch a cold].
We can easily nd more adjectives which select a CP complement:
91
(73) a. I am ashamed that I neglected you.
b. I am delighted that Mary nished his thesis.
c. We are content for the cleaners to return the drapes next week.
d. We were thankful that no one had been hurt.
e. We were glad it was over.
The lexical entries for some adjectives are given in (74):
(74) a.
_
_
condent
HEAD| POS adj
COMPS CP[VFORM n]
_
_
b.
_
_
insistent
HEAD| POS adj
COMPS CP[VFORM bse]
_
_
c.
_
_
eager
HEAD| POS adj
COMPS [VFORM inf ]
_
_
Such lexical entries, interacting with the Head-Complement Rule, the Head-Specier Rule, and
the HFP, can license analyses such as (75), for (72b):
(75)
S
NP
VP
Tom
V
_
COMPS 2
_ 2 AP
seems
A
_
COMPS 3
_
3 VP
_
VFORM inf
_
eager V
VP
_
VFORM bse
_
to V NP
catch a cold
92
When the head adjective eager combines with its complement, VP[inf ], it satises the Head-
Complement Rule. The same rule allows the combination of the verb seem with its AP comple-
ment.
5.4.4 Nouns Selecting a Clausal Complement
Nouns can also select a CP complement, for example, eagerness:
(76) a. (Johns) eagerness [for Harry to win the election]
b. (Johns) eagerness [to win the election]
These examples imply that eagerness will have the following lexical information:
(77)
_
_
eagerness
HEAD| POS noun
COMPS
_
[VFORM inf ]
_
_
_
This lexical entry will allow a structure like the following:
(78)
NP
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
DP
.
.
.
.
.
.
.
.
.
Johns
N
_
COMPS 2
_
2 VP
_
VFORM inf
_
_
my
HEAD| POS det
SPR
COMPS
_
_
b.
_
_
s
HEAD| POS det
SPR NP
COMPS
_
_
c.
_
_
brother
HEAD| POS noun
SPR DP
COMPS
_
_
d.
_
_
friend
HEAD| POS noun
SPR DP
COMPS
_
_
As given here, determiners like my and uncountable nouns like rice select neither specier
nor complement whereas the possessive marker s and the countable noun friend requires a
NP and DP specier, respectively. Meanwhile, countable noun brother requires a DP as its
specier. This means that not only a simple determiner but also a possessive NP can function
as a specier. Reecting this generalization, we can assume that a countable noun selects not a
simple determiner but a DP (determiner phrase) as its specier: These lexical entries will project
NP structures like the following for (5b):
2
(7)
NP
1 DP
_
_
SPR 1 DP
COMPS
_
_
2 NP
Det
_
_
SPR 2 NP
COMPS
_
_
N
_
_
SPR 1 DP
COMPS
_
_
my brother s friend
2
Keen readers may have noticed that we allow a vacuous projection from N to N
_
a
HEAD
_
_
POS det
AGR| NUM sing
_
_
SPR
COMPS
_
_
b.
_
_
book
HEAD
_
_
POS noun
AGR | NUM sing
_
_
SPR DP[NUM sing]
COMPS
_
_
Common nouns thus impose a specic NUM value on the specier:
103
(16)
NP
_
AGR| NUM sing
_
1 DP[NUM sing]
N
_
AGR| NUM sing
SPR 1 DP
COMPS
_
a
N
_
_
AGR| NUM sing
SPR 1 DP
COMPS
_
book
The singular noun book selects a singular determiner like a. Notice that the AGR value on the
head noun book is passed up to the whole NP, marking the whole NP as singular, so that it can
combine with a singular VP, if it is the subject.
In addition, there is nothing preventing a singular noun from combining with a determiner
which is not specied at all for a NUM value:
(17) a. *those book, *these book, . . .
b. no book, the book, my book, . . .
Determiners like the, no and my are not specied for a NUM value. Formally, their NUM value
is underspecied as num(ber). That is, the grammar of English has the underspecied value num
for the feature NUM, with two subtypes sing(ular) and pl(ural):
(18)
num
sing pl
Given this hierarchy, nouns like book requiring a singular Det can combine with determiners
like the whose AGR value is num. This is in accord with the grammar since the value num is a
supertype of sing.
104
6.3.2 Pronoun-Antecedent Agreement
As noted earlier, a second type of agreement is pronoun-antecedent agreement, as indicated in
(19).
(19) a. In the book, he talks about his ups and downs at McLaren. Throughout it all he
seeks to answer the questions about himself.
b. If John wants to succeed in corporate life, he/*she has to know the rules of the
game.
c. The critique of Platos Republic was written from a contemporary point of view. It
was an in-depth analysis of Platos opinions about possible governmental forms.
The pronoun he or it here needs to agree with its antecedent not only with respect to the number
value but also with respect to person (1st, 2nd, 3rd) and gender (masculine, feminine, or neuter)
values too. This shows us that nouns have also information about person and gender as well as
number in the AGR values:
(20)
a.
_
_
book
HEAD
_
_
POS noun
AGR
_
_
NUM sing
GEN neut
PER 3rd
_
_
_
_
VAL
_
_
SPR DP[NUM sing]
COMPS
_
_
_
_
b.
_
_
he
HEAD
_
_
POS noun
AGR
_
_
NUM sing
PER 3rd
GEN masc
_
_
_
_
VAL
_
_
SPR
COMPS
_
_
_
_
As we have briey shown, nouns have NUM, PER(SON), and GEN(DER) for their AGRvalues.
The PER value can be 1st, 2nd or 3rd; the GEN value can be masc(uline), fem(inine) or neut(er).
The NUM values are shown in (18) above.
6.3.3 Subject-Verb Agreement
The third type of agreement is subject-verb agreement, which is one of the most important
phenomena in English syntax. Let us look at some slightly complex examples:
(21) a. The characters in Shakespeares Twelfth Night *lives/live in a world that has been
turned upside-down.
b. Students studying English read/*reads Conrads Heart of Darkness while at uni-
versity.
As we can see here, the subject and the verb need to have an identical number value; and the
person value is also involved in agreement relations, in particular when the subject is a personal
pronoun:
(22) a. You are/*is the only person that I can rely on.
105
b. He is/*are the only person that I can rely on.
These facts show us that a verb lexically species the information about the number as well as
person values of the subject that it selects for.
To show how the agreement system works, we will use some simpler examples:
(23) a. The boy swims/*swim.
b. The boys swim/*swims.
English verbs will have at least the following selectional information:
(24)
_
_
swims
HEAD
_
_
POS verb
VFORM pres
_
_
VAL| SPR
_
NP
_
_
PER 3rd
NUM sing
_
_
_
_
_
The present-tense verb swims thus species that its subject (SPRs value) carries a 3rd singular
AGR information. This lexical information will license a structure like the following:
(25)
S
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2 NP
_
_
AGR
_
_
PER 3rd
NUM sing
_
_
_
_
VP
_
_
SPR
_
2 NP
_
_
AGR
_
_
PER 3rd
NUM sing
_
_
_
_
_
_
_
The boy
V
_
SPR 2 NP
_
swims
Only when the verb combines with a subject satisfying its AGR requirement will we have a
well-formed head-subject phrase. In other words, if this verb were to combine with a subject
with an incompatible agreement value, we would generate an ungrammatical example like *The
boys swims in (23b). In this system, what subject-verb agreement is is structure-sharing between
the AGR value of the subject (SPR value of the verb) and that of the NP that the VP combines
with.
106
The acute reader may have noticed that there are similarities between noun-determiner agree-
ment and subject-verb agreement, that is, in the way that agreement works inside NP and inside
S. Both NP and S require agreement between the head and the specier, as reected in the
revised Head-Specier Rule in (14).
6.4 Semantic Agreement Features
What we have seen so far is that the morphosyntactic AGR values of noun or verb can be spec-
ied, and may be inherited by phrases built out of them. However, consider now the following
examples adopted from Nunberg (1995):
(26) a. [The hash browns at table nine] are/*is getting cold.
b. [The hash browns at table nine] is/*are getting angry.
When (26b) is spoken by a waiter to another waiter, the subject refers to a person who ordered
hash browns. A somewhat similar case is found in (27):
(27) King prawns cooked in chili salt and pepper was very much better, a simple dish
succulently executed.
Here the verb form was is singular to agree with the dish being referred to, rather than with a
plurality of prawns. If we simply assume that the subject phrase inherits the morphosyntactic
agreement features of the head noun (hash) browns in (26b) and (King) prawns in (27), and
requires that these features match those of the verb, we would not expect the singular verb form
to be possible at all in these examples. In the interpretation of a nominal expression, it must
be anchored to an individual in the situation described. We call this anchoring value the noun
phrases index value. The index of hash browns in (26a) must be anchored to the plural entities
on the plate, whereas that of hash browns in (26b) is anchored to a customer who ordered the
food.
English agreement is not purely morpho-syntactic as described in the sections above, but
context-dependent in various ways, via the notion of index that we have just introduced. Often
what a given nominal refers to in the real world is important for agreement index agreement.
Index agreement involves sharing of referential indexes, closely related to the semantics of a
nominal, and somewhat separate from the syntactic agreement feature AGR. This then requires
us to distinguish the morphological AGR value and semantic (SEM(ANTIC)) IND(EX) value.
So, in addition to the morphological AGR value introduced above, each noun will also have a
semantic IND value representing what the noun refers to in the actual world.
(28) a.
_
_
boy
SYN| HEAD
_
_
POS noun
AGR| NUM sing
_
_
SEM| IND| NUM sing
_
_
107
b.
_
_
boys
SYN| HEAD
_
_
POS noun
AGR| NUM pl
_
_
SEM| IND| NUM pl
_
_
The lexical entry for boy indicates that it is syntactically a singular noun (through the feature
AGR) and semantically also denotes a singular entity (through the feature IND). And the verb
will place a restriction on its subjects IND value rather than its morphological AGR value:
4
(29)
_
_
swims
SYN
_
_
HEAD
_
_
POS verb
AGR| NUM sing
_
_
VAL| SPR
_
NP[IND| NUM sing]
_
_
_
SEM| IND s0
_
_
The lexical entry for swims here indicates that it is morphologically marked as singular (the
AGR feature) and selects a subject to be linked to a singular entity in the context (by the feature
IND). Different from nouns, the verbs own IND value is a situation index (s0) in which the in-
dividual referred to through the SPR value is performing the action of swimming. If the referent
of this subject (its IND value) does not match, we would generate an ungrammatical example
like *The boys swims:
(30)
*S
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
2 NP
_
_
IND
j
_
_
PER 3rd
NUM pl
_
_
_
_
VP
_
_
SPR
_
2 NP
i
_
_
PER 3rd
NUM sing
_
_
_
_
_
The boys
V
_
SPR 2 NP
i
_
swims
4
The IND value of a noun will be an individual index (i, j, k, etc) whereas that of a verb or predicative adjective will
be a situation index such as s0, s1, s2, etc.
108
As we can see here, the required subject has the IND value i, but the subject has a different IND
value j.
In the most usual cases, the AGR and IND value are identical, but they can be different, as in
examples like (26b). This means that depending on the context, hash browns can have different
IND values:
5
(31)
a.
_
_
hash browns
SYN| HEAD
_
_
POS noun
AGR| NUM pl
_
_
SEM| IND 1 | NUM pl
_
_
(when referring to the food itself)
b.
_
_
hash browns
SYN| HEAD
_
_
POS noun
AGR| NUM pl
_
_
SEM| IND 1 | NUM sing
_
_
(when referring to a customer, or to a dish)
In the lexical entry (31b), the AGRs NUM value is plural but its INDs NUM value is singu-
lar. As shown by (26), the reference hash browns can be transferred from cooked potatoes to
the customer who ordered it. This means that given an appropriate context, there could be a
mismatch between the morphological form of a noun and the index value that the noun refers
to.
What this indicates is that subject-verb agreement and noun-specier agreement are dif-
ferent. In fact, English determiner-noun agreement is only a reection of morpho-syntactic
agreement features between determiner and noun, whereas both subject-verb agreement and
pronoun-antecedent agreement are index-based agreement. This is represented in (32).
(32) Morpho-syntactic agreement (AGR)
Det
head-noun
verb . . .
Index agreement (IND)
Such agreement patterns can be clearly found in examples like the following where the un-
derlined parts have singular agreement with four pounds, which is internally plural.
(33) [Four pounds] was quite a bit of money in 1950 and it was not easy to come by.
Given the separation of the morphological AGR value and the semantic IND value, nothing
blocks mismatches between the two (AGR and IND) as long as all the other constraints are
satised. Observe further examples in the following:
(34) a. [Five pounds] is/*are a lot of money.
5
As indicated here, the lexical expression now has two features SYN (SYNTAX) and SEM (SEMANTICS). The
feature SYN includes HEAD and SPR and COMPS. The feature SEM is for semantic information. As our discussion
goes on, we will add more to this part.
109
b. [Two drops] deodorizes/*deodorize anything in your house.
c. [Fifteen dollars] in a week is/*are not much.
d. [Fifteen years] represents/*represent a long period of his life.
e. [Two miles] is/*are as far as they can walk.
In all of these examples with measure nouns, the plural subject combines with a singular verb.
An apparent conict arises from the agreement features of the head noun. For proper agreement
inside the noun phrase, the head noun has to be plural, but for subject-verb agreement the
noun has to be singular. Consider the example in (34a). The nouns pounds and drops here are
morphologically plural and thus must select a plural determiner, as argued so far. But when
these nouns are anchored to the group as a whole that is, conceptualized as referring to a
single measure, the index value has to be singular, as represented in (35).
(35)
_
_
pounds
SYN
_
_
HEAD
_
_
POS noun
AGR 1 | NUM pl
_
_
SPR
_
DP
_
AGR 1
_
_
_
_
SEM| IND | NUM sing
_
_
As indicated in the lexical entry (35), the morpho-syntactic number value of pounds is plural
whereas the index value is singular. In the present analysis, this would mean that pounds will
combine with a plural determiner but with a singular verb. This is possible, as noted earlier in
section 2, since the index value is anchored to a singular individual in the context of utterance.
The present analysis thus generates the following structure for the sentence (34a):
(36)
S
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
3 NP
_
_
AGR 1
IND i
_
_
VP
_
SPR 3 NP
i
Det
_
AGR 1 | NUM pl
_
N
_
_
AGR 1
IND i
_
_
V
_
SPR 3 NP
i
_
NP
_
this
HEAD
_
_
POS det
AGR| NUM sing
_
_
_
_
b.
_
_
team/government
SYN
_
_
HEAD
_
_
POS noun
AGR| NUM sing
_
_
_
_
SEM| IND | NUM pl
_
_
As represented in (38a) and (38b), this and government agree each other in terms of the morpho-
syntactic agreement number value whereas the index value of government is what matters for
subject-verb agreement. This in turn means that when government refers to the individuals in
this given group, the whole NP this government carries a plural index value.
6.5 Partitive NPs and Agreement
6.5.1 Basic Properties
With regard to the NP-internal elements between which we may nd instances of agreement,
there are two main types of NP in English: simple NPs and partitive NPs, shown in (39) and
(40) respectively.
(39) a. some objections
b. most students
c. all students
6
To some speakers, the plural verb does not go with the singular subject this team. To these people, the neutral
determiner the is much better.
111
d. much worry
e. many students
f. neither cars
(40) a. some of the objections
b. most of the students
c. all of the students
d. much of her worry
e. many of the students
f. neither of the cars
As in (40), the partitive phrases have a quantier followed by an of -phrase, designating a set
with respect to which certain individuals are quantied. In terms of semantics, these partitive
NPs are different from simple NPs in several respects.
First, the lower NP in partitive phrases must be denite; and in the of -phrase, no quantica-
tional NP is allowed, as shown in (41):
(41) a. Each student vs. each of the students vs. *each of students
b. Some problems vs. some of the problems vs. *some of many problems
Second, not all determiners with quanticational force can appear in partitive constructions.
As shown in (42), determiners such as the, every and no cannot occupy the rst position:
(42) a. *the of the students vs. the students
b. *every of his ideas vs. every idea
c. *no of your books vs. no book(s)
Third, simple NPs and partitive NPs have different restrictions relative to the semantic head.
Observe the contrast between (43) and (44):
(43) a. She doesnt believe much of that story.
b. We listened to as little of his speech as possible.
c. How much of the fresco did the ood damage?
d. I read some of the book.
(44) a. *She doesnt believe much story.
b. *We listened to as little speech as possible.
c. *How much fresco did the ood damage?
d. *I read some book.
The partitive constructions in (43) allow a mass (non-count) quantier such as much, little and
some to cooccur with a lower of -NP containing a singular count noun. But as we can see in
(44), the same elements serving as determiners cannot directly precede such nouns.
Another difference concerns lexical idiosyncrasies.
(45) a. One of the people was dying of thirst.
112
b. Many of the people were dying of thirst.
(46) a. *One people was dying of thirst.
b. Many people were dying of thirst.
The partitives can be headed by quantiers like one and many, as shown in (45) and (46) but
unlike many, one cannot serve as a determiner when the head noun is collective as in (46a).
6.5.2 Two Types of Partitive NPs
We classify partitive NPs into two types based on the agreement facts, and call them Type I
and Type II. In Type I, the number value of the partitive phrase depends on the preceding head
noun whereas in Type II, the number value depends on the head noun inside the of -NP phrase.
Observe Type I examples.
(47) Type I:
a. Each of the suggestions is acceptable.
b. Neither of the cars has air conditioning.
c. None of these men wants to be president.
We can observe here that the verbs number value is determined by the preceding expression
each, neither and none. Now see Type II:
(48) Type II:
a. Most of the fruit is rotten.
b. Most of the children are here.
c. Some of the soup needs more salt.
d. Some of the diners need menus.
e. All of the land belongs to the government.
f. All of these cars belong to me.
As shown in (48), when the NP following the preposition of is singular or uncountable, the main
verb is singular. When the NP is plural, the verb is also plural. From a semantic perspective, we
see that the class of quanticational indenite pronouns including some, half, most and all may
combine either singular or plural verbs, depending upon the reference of the of -NP phrase. If
the meaning of these phrases is about how much of something is meant, the verb is singular; but
if the meaning is about how many of something is meant, the verb is plural. The expressions in
(49) also exhibit similar behavior in terms of agreement.
(49) half of, part of, the majority of, the rest of, two-thirds of, a number of (but not the
number of )
An effective way of capturing the relations between Type I and Type II constructions involves
the lexical properties of the quantiers. First, Type I and Type II involve pronominal forms
serving as the head of the constructions, which select an of -NP inside which the NP is denite:
(50) a. *neither of students, *some of water
113
b. neither of the two linguists/some of the water
However, we know that the two types are different in terms of agreement: the pronouns in the
Type I construction are lexically specied to be singular whereas the number value for Type II
comes from inside the selected PP.
A slight digression is in order. It is easy to see that there are prepositions whose functions
are just grammatical markers.
(51) a. John is in the room.
b. I am fond of him.
The predicative preposition in here selects two arguments John and the room. Meanwhile, the
preposition of has no predicative meaning, but just functions as a marker to the argument of
fond. As for the PPs headed by these markers, as in the partitive construction, their semantic
features are identical with the prepositional object NP. There is no semantic difference (such as
deniteness effect represented as the feature DEF in the present system) between the PP of him
and the NP him.
Given this analysis in which the PP in the partitive construction has the identical AGR and
semantic features (e.g., DEF) with its inner NP, we can lexically encode the similarities and
differences between Type I and Type II in a simple manner:
(52) a.
_
_
neither
HEAD
_
_
POS noun
AGR| NUM sing
_
_
COMPS
_
PP
_
_
PFORM of
DEF +
_
_
_
_
_
b.
_
_
some
HEAD
_
_
POS noun
AGR| NUM 1
_
_
COMPS
_
PP
_
_
PFORM of
DEF +
AGR| NUM 1
_
_
_
_
_
(52) shows that both Type I neither and Type II some are lexically specied to require a PP
complement whose semantic value includes the denite (DEF) feature (with the value +). This
will account for the contrast in (50). However, the two types are different in terms of their
AGRs NUM value. The NUM value of Type I neither is singular, whereas that of Type II is
identied with the PPs NUM value which is actually coming from its prepositional object NP.
Showing these differences in the syntactic structures, we have the alternatives in (53):
114
(53) a.
NP[NUM sing]
N[NUM sing]
PP
neither P
NP
of the students
b.
NP[NUM 1 ]
N[NUM 1 ]
PP[NUM 1 ]
some P
NP[NUM 1 ]
.
of the students
As shown in (53a), for Type I, it is neither which determines the NUM value of the whole
NP phrase. However, for Type II, it is the NP the students which determines the NUM value of
the whole NP.
We can check a few of the consequences of these different specications in the two Types.
Consider the contrast in (54):
(54) a. many of the/those/her apples
b. *many of some/all/no apples
(54b) is ungrammatical since many requires an of -PP phrase whose DEF value is positive.
This system also offers a simple way of dealing with the fact that quantiers like each affect
the NUM value as well as the countability of the of -NP phrase. One difference between Type I
and Type II is that Type I selects a plural of -NP phrase whereas Type II has no such restriction.
This is illustrated in (55) and (56).
(55) Type I:
a. one of the suggestions/*the suggestion/*his advice
b. each of the suggestions/*the suggestion/*his advice
c. neither of the students/*the student/*his advice
(56) Type II:
a. some of his advice/students
b. most of his advice/students
115
c. all of his advice/students
The only additional specication we need for Type I pronouns relates to the NUM value on the
PPs complement as given in (57):
(57)
_
_
each
HEAD
_
_
POS noun
AGR| NUM sing
_
_
COMPS
_
PP
_
_
PFORM of
DEF +
NUM pl
_
_
_
_
_
We see that quantiers like each select a PP complement whose NUM value is plural.
Type II pronouns do not have such a requirement on the PP complement note that all the
examples in (58) are acceptable, in contrast to those in (59) (cf. Baker 1995):
(58) a. Most of Johns boat has been repainted.
b. Some of the record contains evidence of wrongdoing.
c. Much of that theory is unfounded.
(59) a. *Each of Johns boat has been repainted.
b. *Many of the record contained evidence of wrongdoing.
c. *One of the story has appeared in your newspaper.
The contrast here indicates that Type II pronouns can combine with a PP whose inner NP is
singular. This is simply predicted since our analysis allows the inner NP to be either plural or
singular (or uncountable).
We are also in a position now to understand some differences between simple NPs and par-
titive NPs. Consider the following examples:
(60) a. many dogs/*much dog/the dogs
b. much furniture/*many furniture/the furniture
(61) a. few dogs/*few dog/*little dogs/*little dog
b. little furniture/*little furnitures/*few furniture/*few furnitures
The data here indicate that in addition to the agreement features we have seen so far, common
nouns also place a restriction on the countability value of the selected specier. Specically, a
countable noun selects a countable determiner as its specier. To capture this agreement restric-
tion, we introduce a new feature COUNT (COUNTABLE):
116
(62)
a.
_
_
dogs
HEAD| POS noun
SPR DP[COUNT +]
_
_
b.
_
_
furniture
HEAD| POS noun
SPR DP[COUNT ]
_
_
The lexical specication on a countable noun like dogs requires its specier to be [COUNT +],
to prevent formations like *much dogs. This in turn means that determiners must also carry the
feature COUNT:
(63)
a.
_
_
many
HEAD
_
_
POS det
COUNT +
_
_
_
_
b.
_
_
the
HEAD
_
_
POS det
COUNT +/
_
_
_
_
c.
_
_
little
HEAD
_
_
POS det
COUNT
_
_
_
_
Notice here that some determiners such as the are not specied for a value for COUNT. Ef-
fectively, the value can be either + or , licensing combination with either a countable or an
uncountable noun (the book or the furniture).
Now consider the following contrast:
(64) a. much advice vs. *many advice
b. *much story vs. many stories
(65) a. much of the advice vs. *many of the advice
b. much of the story vs. many of the stories
Due to the feature COUNT, we understand now the contrast between much advice and *many
advice or the contrast between *much story and many stories. The facts in partitive structures
are slightly different, as (65) shows, but the patterns in the data directly follow from these lexical
entries:
(66)
a.
_
_
many
HEAD| POS noun
COMPS
_
PP
_
_
PFORM of
NUM pl
DEF +
_
_
_
_
_
b.
_
_
much
HEAD| POS noun
COMPS
_
PP
_
_
PFORM of
NUM sing
DEF +
_
_
_
_
_
The pronoun many requires a PP complement whose inner NP is plural, whereas much does not.
117
6.5.3 Measure Noun Phrases
There are also so-called measure noun phrase constructions, which are similar to partitive
constructions. Consider the following contrast:
(67) a. one pound of those beans
b. three feet of that wire
c. a quart of Bobs cider
(68) a. one pound of beans
b. three feet of wire
c. a quart of cider
Notice here that (67) is a kind of partitive construction whereas (68) just measures the amount
of the NP after of . As the examples show, measure noun phrases do not require a denite article,
which is not an option for the true partitive constructions, repeated here:
(69) *many of beans, *some of wire, *much of cider, *none of yogurt, *one of straw-
berries
In addition, there are several more differences between partitive and measure noun phrases.
For example, measure nouns cannot occur in simple noun phrases. They obligatorily require an
of -NP phrase:
(70) a. *one pound beans vs. one pound of beans
b. *three feet wire vs. three feet of wire
c. *a quart cider vs. a quart of cider
Further, unlike partitive constructions, measure noun phrases require a numeral as their spec-
ier:
(71) a. *one many of the books, *several much of the beer
b. one pound of beans, three feet of wire
As noted here, many or much in the partitive constructions cannot combine with numerals like
one or several; measure nouns pound and feet need to combine with a numeral like one or three.
Further complications arise due to the existence of defective measure noun phrases. Consider
the following examples:
(72) a. *a can tomatoes/a can of tomatoes/one can of tomatoes
b. a few suggestions/*a few of suggestions/*one few of suggestions
c. *a lot suggestions/a lot of suggestions/*one lot of suggestions
Expressions like few and lot actually behave quite differently. With respect to few, it appears
that a few acts like a complex word. However, lot acts more like a noun, but unlike can, it does
not allow its specier to be a numeral.
In terms of agreement, measure noun phrases behave like Type I partitive constructions:
(73) a. A can of tomatoes is/*are added.
118
b. Two cans of tomatoes are/*is added.
We can see here that it is the head noun can or cans which determines the NUM value of the
whole NP. The inner NP in the PP does not affect the NUM value at all. These observations lead
us to posit the following lexical entry for a measure noun:
(74)
_
_
pound
HEAD
_
_
POS noun
NUM sing
_
_
SPR DP
COMPS
_
PP
_
PFORM of
_
_
_
_
That is, a measure noun like pound requires one obligatory SPR and a PP complement. Unlike
partitive constructions, there is no deniteness restriction on the PP complement.
Finally, there is one set of words whose behavior leaves them somewhere between quantity
words and measure nouns. These are words such as dozen, hundred, and thousand:
(75) a. three hundred of your friends
b. *three hundreds of your friends
c. *three hundreds of friends
d. three hundred friends
e. hundreds of friends/*hundreds friends
Consider the behavior of hundred and hundreds here. The singular hundred, when used as
noun, obligatorily requires a PP[of ] complement as well as a numeral specier, as in (75a). The
plural hundreds requires no specier although it also selects a PP complement. Not surprisingly,
similar behavior can be observed with thousand and thousands:
(76) a. several thousand of Bills supporters
b. *several thousands of Bills supporters
c. *several thousands of supporters
d. several thousand supporters
e. thousands of supporters/*thousands supporters
One way to capture these properties is to assign the following lexical specications to hundred
and hundreds:
(77)
a.
_
_
hundred
HEAD| POS noun
SPR DP
COMPS
_
PP[PFORM of]
_
_
_
b.
_
_
hundreds
HEAD| POS noun
SPR
COMPS
_
PP[PFOR of]
_
_
_
119
Even though there may be some semantic reasons for all these different kinds of lexical spec-
ications, for now, stating it all directly in the lexical entries will account at least for the data
given here.
6.6 Modifying an NP
6.6.1 Adjectives as Prenominal Modiers
Adjectives are expressions commonly used to modify a noun. However, not all adjectives can
modify nouns. Even though most adjectives can be used either as in a modifying (attributive)
function or as a predicate (as in She is tall), certain adjectives are restricted to their usages.
Adjectives such alive, asleep, awake, afraid, ashamed, aware, can be used only predicatively,
whereas others such as wooden, drunken, golden, main and mere are only used attributively:
(78) a. He is alive.
b. He is afraid of foxes.
(79) a. It is a wooden desk.
b. It is a golden hair.
c. It is the main street.
(80) a. *It is an alive sh. (cf. living sh)
b. *They are afraid people. (cf. nervous people)
(81) a. *This objection is main. (cf. the main objection)
b. *This fact is key. (cf. a key fact)
The predicatively-used adjectives are specied with the feature PRD, and with a MOD value
being empty as default, as shown here:
7
(82)
_
_
alive
HEAD
_
_
POS adj
PRD +
MOD
_
_
_
_
This says that alive is used predicatively, and does not have a specication for a MOD value
(the value is empty). This lexical information will prevent predicative adjectives from also func-
tioning as noun modiers.
8
In contrast to the predicative adjective, a modifying adjective will have the following lexical
entry:
7
All modiers (adverbial element) will carry the head feature MOD.
8
In addition, all predicative expressions select one argument, their subject (SPR). This information is not shown
here.
120
(83)
_
_
brave
HEAD
_
_
POS adj
MOD N
_
_
_
_
This species an adjective which modies any value whose POS is noun. This will license a
structure like the following:
(84)
NP
_
_
_
_
_
_
_
_
_
2 DP
N
_
SPR 2 DP
_
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
the
AP
_
MOD 1
_
1 N
_
SPR 2 DP
_
brave child
6.6.2 Postnominal Modiers
Postnominal modiers are basically the same as prenominal modiers with respect to what they
are modifying. The only difference is that they come after what they modify. Various phrases
can function as such postnominal modiers:
(85) a. [The boy [in the doorway]] waved to his father.
b. [The man [eager to start the meeting]] is Johns sister.
c. [The man [holding the bottle]] disappeared.
d. [The papers [removed from the safe]] have not been found.
e. [The money [that you gave me]] disappeared last night.
All these postnominal elements bear the feature MOD. Leaving aside detailed discussion of the
relative clause(-like) modiers in be until Chapter 12, we can say that example (85a) will have
the following structure:
121
(86)
NP
_
_
_
_
_
_
_
_
_
2 DP
N
_
SPR 2 DP
_
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
the
1 N
_
SPR 2 DP
_
PP
_
MOD 1
_
.
.
.
.
.
.
.
NP
VP
.
.
.
.
.
.
.
.
.
.
.
V S
.
.
.
.
.
.
.
.
.
.
.
seems NP VP[inf ]
Stephen
to be irritating
The movement of the subject Stephen to the higher subject position will correctly generate
(26b). This kind of movement to the subject position can be triggered by the requirement that
each English declarative have a surface subject (cf. Chomsky 1981). A similar movement pro-
cess can be applied to the object raising cases:
(28) a. Deep structure: Tom believes [Stephen to be irritating].
b. Surface structure: Tom believes Stephen to be irritating.
Here the embedded subject Stephen moves not to the matrix subject but to the matrix object
position:
132
(29)
S
NP
VP
John
V
NP
S
believes NP VP[inf ]
Mary
to be irritating
Control constructions are different: there is no movement operation involved. Instead, it is
the lower subject position which has special properties. Consider the examples in (30):
(30) a. John tried to please Stephen.
b. John persuaded Stephen to be more careful.
Since try and persuade assign a semantic roles to their subject, and objects, an unlled position
of the kind designated above by cannot be allowed. Instead, it is posited that there is an
unexpressed subject of the innitival VP to please Stephen and to be more careful. This is
traditionally represented as the element called PRO (a silent pronoun), and the examples
will have the following deep structures:
(31) a. John tried [PRO to please Stephen].
b. John persuaded Stephen [PRO to be more careful].
The nal tree representations of these are as follows:
(32) a.
S
NP
VP
John
i
V S
tried NP VP[inf ]
PRO
i
to please Stephen
133
b.
S
NP
VP
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
John
V
NP
S
persuaded Stephen
i
NP VP[inf ]
PRO
i
to be more careful
An independent part of the theory of control links PRO in each case to its antecedent, marked
by coindexing. In (32a), PRO is coindexed with John whereas in (32b), it is coindexed with
Stephen.
These analyses which involve derivations on tree structures are driven by the assumption
that the mapping between semantics and syntax is very direct. For example, in (29), the verb
believe semantically selects an experiencer and a proposition, and this is reected in the initial
structure. In some syntactic respects, though, believe acts like it has an NP object (separate from
the inntival complement), and the raising operation creates this object. In contrast, persuade
semantically selects an agent, a patient, and a proposition, and hence the structure in (32b)
reects this: the object position is there all along, so to speak.
The classical transformational approach provides a useful graphical approach to understand-
ing the difference between raising and control. However, it requires assumptions about the na-
ture of grammar rather different from what we have made throughout this book. In the rest of
this chapter, we present a nontransformational account of control and raising.
7.4 A Nontransformational Approach
7.4.1 Identical Syntactic Structures
Instead of the movement approach in which movement operations and various kinds of empty
elements or positions play crucial roles, we simply focus directly on the surface structures of
raising and control constructions. Going back to seem and try, we can observe that both select
an innitival VP, as in (33), giving the structures in (34):
(33) a.
_
_
seems
SPR NP
COMPS
_
VP
_
VFORM inf
_
_
_
_
134
b.
_
_
tries
SPR NP
COMPS VP
_
VFORM inf
_
_
_
(34) a.
S
.
.
.
.
.
.
.
.
.
.
.
NP
VP
.
.
.
.
.
.
.
.
.
.
.
John
V
VP[inf ]
.
.
.
.
.
.
.
.
.
.
.
seems V VP[bse]
to be honest
b.
S
.
.
.
.
.
.
.
.
.
.
.
NP
VP
.
.
.
.
.
.
.
.
.
.
.
John
V
VP[inf ]
.
.
.
.
.
.
.
.
.
.
.
tries V VP[bse]
to be honest
As shown here, seems and tries actually have identical structures.
The object raising verb expect and the control verb persuade also have identical valence
(SPR and COMPS) information:
(35) a.
_
_
expects
SPR NP
COMPS NP, VP
_
VFORM inf
_
_
b.
_
_
persuaded
SPR NP
COMPS NP, VP
_
VFORM inf
_
_
These two lexical entries will license the following structures:
135
(36) a.
S
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
NP
VP
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
Kim V
NP
VP
VP
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
Kim V
NP
VP
_
seemed
SPR 1
COMPS
_
VP
_
_
VFORM inf
SPR 1
_
_
_
_
_
b.
_
_
tried
SPR NP
i
COMPS
_
VP
_
_
VFORM inf
SPR NP
i
_
_
_
_
_
These two lexical entries represent the difference between seem and try: for seemed, the subject
of the VP complement is identical with its own subject (notated by 1 ) whereas for tried, only the
index value of its VP complement is identical to that of its subject. That is, the VP complements
understood subject refers to the same individual as the subject of tried. This index identity in
control constructions is clear when we consider examples like the following:
(40) Someone
i
tried NP
i
to leave the town.
The example here means that whoever someone might refer to, that same person left town.
The lexical entries in (39) generate following structures for the intransitive raising and control
sentences:
137
(41) a.
S
_
_
HEAD 4 | POS verb
SPR
COMPS
_
_
_
_
_
_
_
_
_
_
_
_
_
1 NP
VP
_
_
HEAD 4
SPR 1
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
John
V
_
_
HEAD 4
SPR 1
COMPS 2
_
_
2 VP
_
SPR
_
1
_
_
seems V VP[bse]
to be honest
b.
S
_
_
HEAD 3 | POS verb
SPR
COMPS
_
_
_
_
_
_
_
_
_
_
_
_
_
1 NP
i
VP
_
_
HEAD 3
SPR 1 NP
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
John
V
_
_
HEAD 3
SPR 1 NP
COMPS 2
_
_
2 VP
_
SPR
_
NP
i
_
_
tries V VP[bse]
to be honest
138
It is easy to verify that these structures conform to all the grammar rules (the Head-Specier
Rule and Head-Complement Rule) and principles such as the HFP and the VALP.
Object raising and control predicates are no different. Raising verbs select a VP complement
whose subject is fully identical with the object. Control verbs select a VP complement whose
subjects index value is identical with that of its object. The following lexical entries show these
properties:
(42) a.
_
_
expect
SPR 1 NP
i
COMPS
_
2 NP , VP
_
_
VFORM inf
SPR 2 NP
_
_
_
_
_
b.
_
_
persuade
SPR NP
COMPS
_
NP
i
, VP
_
_
VFORM inf
SPR NP
i
_
_
_
_
_
Let us look at the structures these lexical entries eventually project:
(43)
S
_
_
HEAD 4 | POS verb
SPR
COMPS
_
_
_
_
_
_
_
_
_
_
_
_
_
_
1 NP
VP
_
_
HEAD 4
SPR 1 NP
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
Kim
V
_
_
HEAD 4
SPR 1 NP
COMPS 2 , 3
_
_
2 NP
3 VP
_
SPR
_
2
_
_
_
HEAD 4 | POS verb
SPR
COMPS
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
1 NP
VP
_
_
HEAD 4
SPR 1
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Kim
V
_
_
HEAD 4
SPR 1
COMPS 2 NP, 3 VP
_
_
2 NP
i
3 VP
_
SPR NP
i
_
(j, b)
This shows that the verb hit takes two arguments with the predicate relation hit, with the
notation to indicate the semantic value. The relevant semantic properties can be represented in
a feature structure system as follows:
140
(46)
_
_
hit
SYN| VAL
_
_
SPR NP
i
COMPS NP
j
_
_
SEM
_
_
IND s0
RELS
_
_
_
PRED hit
AGENT i
PATIENT j
_
_
_
_
_
_
_
In terms of syntax, hit is a verb selecting a subject and a complement, as shown in the value
of the feature SYN(TAX). The semantic information of the verb is represented with the fea-
ture SEM(ANTICS). It rst has the attribute IND(EX), representing what this expression refers
to; as a verb, hit refers to a situation s0 in which an individual i hits an individual j. The se-
mantic relation of hitting is represented using the feature for semantic relations (RELS). The
feature RELS has as its value a list of one feature structure, here with three further features,
PRED(ICATE), AGENT, and PATIENT. The predicate (PRED) relation is whatever the verb
denotes: in this case, hit takes two arguments. The AGENT argument in the SEM value is coin-
dexed with the SPR in the SYN value, while the the PATIENT is coindexed with COMPS. This
coindexing links the subcategorization information of hit with the arguments in its semantic
relation. Simply put, the lexical entry in (46) is the formal representation of the fact that in X
hits Y, X is the hitter and Y is the one hit.
Now we can use these extra parts of the representation for the semantic differences in raising
and control verbs. The subject of a raising verb like seem is not assigned any semantic role,
while that of a control verb like try is denitely linked to a semantic role. Assuming that s0
or s1 stands for situations denoted by an innitival VP, seem and try will have the following
simplied meaning representations:
(47) a. seem
_
seem
SYN| VAL
_
_
SPR 1
COMPS
_
VP
_
_
VFORM inf
SPR
_
1
_
IND s1
_
_
_
_
_
SEM
_
_
IND s0
RELS
__
PRED seem
SIT s1
__
_
_
_
_
b.
_
_
try
SYN| VAL
_
_
SPR NP
i
COMPS
_
VP
_
_
VFORM inf
SPR
_
NP
i
_
IND s1
_
_
_
_
_
SEM
_
_
IND s0
RELS
_
_
_
PRED try
AGENT i
SIT s1
_
_
_
_
_
_
_
We can see here that even though the verb seem selects two syntactic arguments, its meaning
relation has only one argument: note that the subject (SPR) is not coindexed with any argument
in the semantic relation. This means that the subject does not receive a semantic role (from
seem). Meanwhile, try is different. Its SPR is coindexed with the AGENT role in the semantics,
and the SPR is also coindexed with the VP complements SPR.
Now we look at object-related verbs like expect and persuade. Just like the contrast between
seem and try, the key difference lies in whether the object (y) receives a semantic role or not:
(49) a. expect
(x, s1)
b. persuade
(x, y, s1)
What one expects, as an experiencer, is a proposition denoted by the VP complement, whereas
what a person x persuades is not a proposition but rather, x persuades an individual y denoted
by the object to perform the proposition denoted by the VP complement. Once again, these
differences are more clearly represented in feature structures:
142
(50) a.
_
_
expect
SYN| VAL
_
_
SPR NP
i
COMPS
_
2 , VP
_
_
VFORM inf
SPR 2 NP
IND s1
_
_
_
_
_
SEM
_
_
IND s0
RELS
_
_
_
PRED expect
EXPERIENCER i
SIT s1
_
_
_
_
_
_
_
b.
_
_
persuade
SYN| VAL
_
_
SPR NP
i
COMPS
_
NP
j
, VP
_
_
VFORM inf
SPR NP
j
IND s1
_
_
_
_
_
SEM
_
_
IND s0
RELS
_
_
_
PRED persuade
AGENT i
THEME j
SIT s1
_
_
_
_
_
_
_
As seen in the lexical entries, expect has two semantic arguments, EXPERIENCER and SIT: the
object is not linked to a semantic argument of expect. In contrast, persuade has three semantic
arguments: AGENT, THEME, and SIT. We can thus conclude that raising predicates assign one
less semantic role in their argument structures than the number of syntactic dependents, while
with control predicates, there is a one-to-one correlation.
7.5 Explaining the Differences
7.5.1 Expletive Subject and Object
Recall that for raising verbs like seem and believe, the subject and object respectively is depen-
dent for its semantic properties solely upon the type of VP complement. This fact is borne out
by the examples in (51):
(51) a. There/*It/*John seems [to be a fountain in the park].
143
b. We believed there/*it/*John [to be a fountain in the park].
Control verbs are different, directly assigning a semantic role to the subject or object. Hence
expletives cannot appear (illustrated here for the subject of try):
(52) a. *There/*It/John tries to leave the country.
b. We believed *there/*it/John to try to leave the country.
7.5.2 Meaning Preservation
We noted above that in a raising example such as (53a), the idiomatic reading can be preserved,
but not in a control example like (53b):
(53) a. The cat seems to be out of the bag.
b. The cat tries to be out of the bag.
This is once again because the subject of seems does not have any semantic role: its subject is
identical with the subject of its VP complement to be out of the bag, whereas the subject of tries
has its own agent role.
Exactly the same explanation applies to the following contrast:
(54) a. The dentist is likely to examine Pat.
b. Pat is likely to be examined by the dentist.
Since likely is a raising predicate, as long as the expressions The dentist examines Pat and Pat
is examined by the dentist have roughly the same meaning, the two raising examples will also
have roughly the same meaning.
However, control examples are different:
(55) a. The dentist is eager to examine Pat.
b. Pat is eager to be examined by the dentist.
The control adjective eager assigns a semantic role to its subject independent of the VP com-
plement, as given in the following lexical entry:
(56)
_
_
eager
SYN| VAL
_
_
SPR NP
i
COMPS
_
VP
_
_
VFORM inf
IND s1
_
_
_
_
_
SEM
_
_
IND s0
RELS
_
_
_
PRED eager
EXPERIENCER i
SIT s1
_
_
_
_
_
_
_
144
This then means that (55a) and (55b) must differ in that in the former, it is the dentist who is
eager to perform the action denoted by the VP complement, whereas in the latter, it is Pat who
is eager.
7.5.3 Subject vs. Object Control Verbs
Consider nally the following two examples:
(57) a. They persuaded me to leave.
b. They promised me to leave.
Both persuaded and promised are control verbs since their object is assigned a semantic role
(and so is their subject). This in turn means that their object cannot be an expletive:
(58) a. *They persuaded it to rain.
b. *They promised it to rain.
However, the two are different with respect to the controller of the innitival VP. Consider who
is understood as the unexpressed subject of the innitival verb here. In (57a), it is the object
me which semantically functions as the subject of the innitival VP. Yet, in (57b), it is the
subject they who will do the action of leaving. Due to this fact, verbs like promise are known as
subject control verbs, whereas those like persuade are object control verbs. This difference
is straightforwardly represented in their lexical entries:
(59)
_
_
persuade
SPR NP
i
COMPS
_
NP
j
, VP
_
_
VFORM inf
SPR NP
j
_
_
_
_
_
_
_
promise
SPR NP
i
COMPS
_
NP
j
, VP
_
_
VFORM inf
SPR NP
i
IND s1
_
_
_
_
_
Based on world knowledge, we know that when one promises someone to do something, this
means that the person who makes the promise will do the action. Meanwhile, when one per-
suades someone to do something, the person who is persuaded will do the action. The lexical
entries here reect this knowledge of the relations in the world.
In sum, the properties of rasing and control verbs presented here can be summarized as
follows:
145
.
Unlike control predicates, raising predicates are unusual in that they do not assign a semantic
role to their subject or object. The absence of a semantic role accounts for the possibility of
expletives it or there or parts of idioms as subject or object with raising predicates, and not
with control predicates.
.
With control predicates, the VP complements unexpressed subject is coindexed with one
of the syntactic dependents. With raising predicates, the entire syntactic-semantic value of
the subject of the innitival VP is structure-shared with that of one of the dependents of
the predicate. This ensures that whatever category is required by the raising predicates VP
complement is the raising predicates subject (or object). Notice that even non-NPs can be
subject in certain kinds of example (see (60)).
(60) a. Under the bed is a fun place to hide.
b. Under the bed seems to be a fun place to hide.
c. *Under the bed wants to be a fun place to hide. (want is a control verb)
146
7.6 Exercises
1. Draw trees for the following sentences and provide the lexical entries for the italicized
verbs:
(i) a. Kim may have admitted to let Mary mow the lawn.
b. Gregory appears to have wanted to be loyal to the company.
c. Jones would prefer for it to be clear to Barry that the city plans to sue him.
d. John continues to avoid the conict.
e. The captain ordered the troops to proceed.
f. He coaxed his brother to give him the candy.
g. Frank hopes to persuade Harry to make the cook wash the dishes.
h. John wants it to be clear to Ben that the city plans to honor him.
2. Explain why the following examples are ungrammatical, based on the lexical entries of
the predicates:
(i) a. *John seems to rain.
b. *John is likely to appear that he will win the game.
c. *Beth tried for Bill to ask a question.
d. *He believed there to be likely that he won the game.
e. *It is likely to seem to be arrogant.
f. *Sandy appears that Kim is happy.
g. *Dana would be unlikely for Pat to be called upon.
h. *Robin is nothing in the box.
i. *It said that Kim was happy.
j. *There preferred for Sandy to get the job.
3. In this chapter, we have learned that predicates (verbs and adjectives) can be classied
into two main groups, raising and control, as represented in the following simple table:
Raising predicates Control predicates
Intransitive seem, . . . try, . . .
Transitive believe, . . . persuade, . . .
Decide in which group each of the following lexical items belongs. In doing so, consider
the it, there, and idiom tests that this chapter has introduced:
(i) certain, anxious, lucky, sure, apt, liable, bound, careful, reluctant
(ii) tend, decide, manage, fail, happen, begin, hope, intend, refuse
4. As we have seen in Exercise 4 of Chapter 3, there is agreement between the copula be
and the postcopular NP in so-called there constructions, as shown again here:
(i) a. There is/*are only one chemical substance involved in nerve transmission.
b. There *is/are more chemical substances involved in nerve transmission.
This kind of agreement relationship can be encoded as a property of the be used in this
construction:
147
(ii)
_
_
be
HEAD
_
_
POS verb
AUX +
_
_
SUBJ
_
NP
_
_
NFORM there
AGR 1
_
_
_
COMPS
_
NP[AGR 1 ], XP[PRED +]
_
_
_
This lexical information species that be selects there as its subject and two complements
(NP and XP). In this case, the complement NPs agreement feature AGR is identical with
that of the subject there. This then will ensure that the verb agrees with the postcopular
NP.
Given this, how might we account for the following contrasts? Provide a structure for
each example and explain the rules or principles which are violated in the ungrammatical
versions:
(ii) a. There is/*are believed to be a sheep in the park.
b. There *is/are believed to be sheep in the park.
c. There seems/*seem to be no student absent.
d. There is/*are likely to be no student absent.
5. Discuss the similarities and differences among the following three examples; use the it,
there and idiom tests.
(i) a. Pat expected Leslie to be aggressive.
b. Pat persuaded Leslie to be aggressive.
c. Pat promised Leslie to be aggressive.
Also, state see the controller is of the innitival VP in each case.
6. Consider the following data and discuss briey what can be the antecedent of her and
herself .
(i) a. Kevin urged Anne to be loyal to her.
b. Kevin urged Anne to be loyal to herself.
Now consider the following data and discuss the binding conditions of ourselves and us.
In particular, determine the relevance of the ARG-ST list with respect to the possible and
impossible binding relations.
(ii) a. We
i
expect the dentist to examine us
i
.
b. *We
i
expect the dentist to examine ourselves
i
.
c. We expect them to examine themselves.
d. *We expect them
i
to examine them
i
.
(iii) a. We
i
persuaded the dentist to examine us
i
.
b. *We
i
persuaded the dentist to examine ourselves
i
.
148
c. We persuaded them
i
to examine themselves
i
.
d. *We persuaded them
i
to examine them
i
.
7. Read the following passage and provide a tree structure for each sentence, and lexical
entries for the underlined words.
(i) If youve ever tried to persuade other people to buy your product or service,
you also know that this can be one of the most discouraging and difcult
things to try to do as a business owner. In fact, this way of trying to get
business by trying to persuade other people is one of the factors that causes
most business owners to dislike or even hate the process of marketing and
selling. Its very tough to try to convince other people to buy from you,
especially if its against their will. After all, if you try to persuade someone
to buy from you, you try to cause that person to do something. And usually
theres always some kind of pressure involved in this process.
149
8
Auxiliary Constructions
8.1 Basic Issues
The English auxiliary system involves a relatively small number of elements interacting with
each other in complex and intriguing ways. This has been one of the main reasons that the
system has been one of the most extensively analyzed empirical domains in the literature on
generative syntax.
Ontological Issues: One of the main issues in the study of English auxiliary system concerns
ontological issues: is it necessary to posit auxiliary as an independent part of speech or not?
Auxiliary verbs can be generally classied as follows:
.
modal auxiliary verbs such as will, shall, may, etc.: have only nite forms
.
have/be: have both nite & nonnite forms
.
do: has a nite form only with vacuous semantic meaning
.
to: has a nonnite form only with apparently vacuous semantic meaning
Such auxiliary verbs behave differently from main verbs in various respects. There have been
arguments for treating these auxiliary verbs as simply having the lexical category V, though
being different in terms of syntactic distribution and semantic contribution. For example, in
terms of similarities, both auxiliary and main verbs behave alike in carrying tense information
and participating in some identical syntactic constructions such as gapping, as shown in (1):
(1) a. John drank water and Bill wine.
b. John may drink water, and Bill drink beer.
Such phenomena provide apparent stumbling blocks to assigning a totally different lexical cat-
egory to English auxiliary verbs, compared to main verbs.
Distinctions between auxiliary and main verbs: One important issue that comes up in the
study of the English auxiliary system is that of which words function as auxiliary verbs, and how
we can differentiate them. Most reliable criteria for auxiliaryhood lie in syntactic phenomena
such as negation, inversion, contraction, and ellipsis (usually known as the NICE properties):
151
1. Negation: Only auxiliary verbs can be followed by not marking sentential negation (in-
cluding have and be):
(2) a. Tom will not leave.
b. *Tom kicked not a ball.
2. Inversion: Only auxiliary verbs undergo subject-auxiliary inversion.
(3) a. Will Tom leave the party now?
b. *Left Tom the party already?
3. Contraction: Only auxiliary verbs have contracted forms with the sufx nt.
(4) a. John couldnt leave the party.
b. *John leftnt the party early.
4. Ellipsis: The complement of an auxiliary verb, but not of a main verb, can be elided.
(5) a. If anybody is spoiling the children, John is .
b. *If anybody keeps spoiling the children, John keeps .
In addition to these NICE properties, tag questions are another criterion: an auxiliary verb can
appear in the tag part of a tag question, but not a main verb:
(6) a. You should leave, shouldnt you?
b. *You didnt leave, left you?
The position of adverbs or so-called oated quantiers can also be adopted in differentiating
auxiliary verbs from main verbs. The difference can easily be seen in the following contrasts:
(7) a. She would never believe that story.
b. *She believed never his story.
(8) a. The boys will all be there.
b. *Our team played all well.
Adverbs such as never and oated quantiers such as all can follow an auxiliary verb, but not a
main verb.
Ordering Restrictions: The third main issue in the syntax of auxiliaries centers on how
to capture the ordering restrictions among them. They are subject to restrictions which limit
the sequences in which they can occur, and the forms in which they can combine with other
auxiliary verbs. Observe the following examples:
(9) a. The children will have been being entertained.
b. He must have been being interrogated by the police at that very moment.
(10) a. *The house is been remodelling.
b. *Margaret has had already left.
c. *He has will seeing his children.
152
d. *He has been must being interrogated by the police at that very moment.
As shown here, when there are two or more auxiliary verbs, they must come in a certain order.
In addition, note that each auxiliary verb requires the immediately following to be in a particular
morphological form (e.g., has eaten vs. *has eating).
In the study of the English auxiliary system, we thus need to address at least the following
issues:
.
Should we posit an auxiliary category?
.
How can we distinguish main verbs from auxiliary verbs?
.
How can we account for phenomena (such as the NICE group) which are sensitive to
the presence of an auxiliary verb?
.
Howcan we capture the ordering and co-occurrence restrictions among auxiliary verbs?
This chapter provides answers to these fundamental questions related to the English auxiliary
system.
1
8.2 Transformational Analyses
The seminal work on the issues above is that of Chomsky (1957). His analysis, introducing the
rule in (11), directly stipulates the ordering relations among auxiliary verbs:
(11) Aux Tense (Modal) (have + en) (be + ing)
The PS rule in (11) would generate sentences with or without auxiliary verbs as in (12):
(12) a. Mary solved the problem.
b. Mary would solve the problem.
c. Mary was solving the problem.
d. Mary would easily solve the problem.
For example, the following structure schematizes some examples in (12):
(13)
S
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
NP
AUX
VP
easily V NP
-theory to provide IP
and CP as categories for clausal syntax, which can deal with the coordination data just given.
Nevertheless, there are many problems which transformational analyses cannot easily overcome
(for a thorough review, see Kim 2000, Kim and Sag 2002).
8.3 A Lexicalist Analysis
In the approach we take in this book, ordering restrictions on auxiliay verbs will follow from
the correct specication of their lexical properties, interacting with the regular rules of syntactic
combination. The analysis requires no movement, either of whole words, or of afxes. In this
section, we discuss several different subtypes of auxiliary.
8.3.1 Modals
One main property of modal auxiliaries such as will, shall and must is that they place no seman-
tic restrictions on their subject, indicating their status as raising verbs (see Chapter 7).
154
(16) a. There might be a unicorn in the garden.
b. It will rain tomorrow.
c. John will leave the party earlier.
(17) a. *There hopes to nish the project.
b. *The bus hopes to be here at ve.
As seen from the contrast, the type of subject in (16) depends on what kind of subject (there or
it or a regular NP) is required by the verb right after the modal. This is typical of raising verbs,
and different from what we see in examples with a control verb like hope in (17), which must
have a referential subject.
Modal verbs can only occur in nite (plain or past) forms. They cannot occur neither as
innitives nor as participles.
2
(18) a. I hope *to would/*to can/to study in France.
b. *John stopped can/canning to sign in tune.
Modals do not show 3rd person inection in the present tense, nor a transparent past tense form.
(19) a. *John musts/musted leave the party early.
b. *John wills leave the party early.
In terms of their own selectional properties, modal verbs select a base VP as their complement:
(20) a. John can [kick/*kicked/*kicking/*to kick the ball].
b. John will [kick/*kicked/*kicking/*to kick the ball].
Reecting these basic lexical properties, a modal auxiliary will have at least the following lexi-
cal information:
(21)
_
_
must
HEAD
_
_
POS verb
VFORM n
AUX +
_
_
VAL
_
_
SPR 1 NP
COMPS
_
VP
_
_
VFORM bse
SPR 1 NP
_
_
_
_
_
_
_
In the lexical information given here, we notice at least three things: rst, auxiliary verbs have
the head feature AUX, which differentiates them from main verbs. In addition, the rule shows
that a modal verb selects a base VP as its complement. This subcategorization information will
rule out examples like the following, as well as the ungrammatical examples in (20):
2
As we have seen in 5.2.1, the VFORM value n includes pres, pst, and pln whereas nonn includes ing, en, inf ,
and bse.
155
(22) a. *Kim must [
VP[n]
bakes a cake].
b. *Kim must [
VP[n]
baked a cake].
c. *Kim must [
VP[n]
will bake a cake].
The possible and impossible structures can be more clearly represented in tree format:
(23) a.
VP
V
_
_
AUX +
COMPS 2 VP[bse]
_
_ 2 VP[bse]
must
V[bse]
NP
bake a cake
b.
*VP
_
_
AUX +
COMPS VP[bse]
_
_ VP[n]
must
V[n]
NP
baked a cake
As can be easily seen here, the modal auxiliary must requires a VP[bse] as its complement:
The VP[n] in (23b) cannot function as the complement of must.
The lexical entry in (21) also species that the VPs subject is identical with the subject of the
modal auxiliary (indicated by the box 1 ). This specication, a crucial property of raising verbs
we have discussed in the previous chapter, rules out the ungrammatical versions of examples
like the following:
(24) a. It/*John will [
VP[bse]
rain tomorrow].
b. There/*It may [
VP[bse]
exist a man in the park].
The VP rain tomorrow requires the expletive subject it, not any other NP, such as John; and the
VP exist a man in the park requires there and nothing else its subject.
In addition, since modal verbs have the specication [VFORM n], they cannot occur in any
environment where nite verbs are prohibited.
156
(25) a. *We expect there to [
VP[n]
will rain].
b. *It is vital that we [
VP[n]
will study everyday].
The simple lexical information of modal verbs given in (21), which is required in almost any
analysis, explains the main distributional possibilities of modal verbs.
8.3.2 Be and Have
The auxiliary verbs have and be are different from modal verbs. For example, unlike modals,
they have nonnite forms (would have, would be, to have/to be); they have a 3rd person inec-
tion form (has, is); they select not a base VP as their complement, but an inected nonnite
form. In addition, they are different from modals in that they also have uses as main verbs,
though in some cases with different syntax from when they are auxiliaries.
Consider the examples in (26):
(26) a. He is a fool.
b. He has a car.
On the assumption that every sentence has a main verb, be and have here are main verbs. How-
ever, a striking property of be is that it still shows the properties of an auxiliary: it exhibits all
of the NICE properties, as we will see below. The usage of be actually provides a strong reason
why the grammar should allow a verb categorized as V to also have the feature specication
[AUX +]; be in (26a) is clearly a verb, yet it also behaves exactly like an auxiliary.
The verb be has three main uses: as a copula selecting an predicate XP, as an aspectual
auxiliary with a progressive VP following, and as an auxiliary as part of the passive construction:
(27) a. John is in the school.
b. John is running to the car.
c. John was found in the ofce.
There is no categorical or syntactic reason to distinguish these three, for they all have NICE
properties: they show identical behavior with subject-auxiliary inversion, their position relative
adverbs including oated quantiers, and so forth.
(28) Subject-Aux Inversion:
a. Was the child in the school? (*Did the child be in the school?)
b. Was the child running to the car?
c. Was the child found?
(29) Position of an adverb:
a. The child (?never) was (never) crazy. (The child (never) became (*never) crazy.)
b. The child (?never) was (never) running to the car.
c. The child (?never) was (never) deceived.
Thus, all three uses have the lexical information given in (30) as their common denominator:
3
3
XP here is a variable over phrasal categories such as NP, VP, AP, and PP.
157
(30)
_
_
HEAD
_
_
POS verb
AUX +
_
_
VAL
_
_
SPR 1 NP
COMPS
_
XP
_
_
PRD +
SPR 1
_
_
_
_
_
_
_
All three bes bear the feature AUX with the + value, and select a predicative phrase whose
subject is identical with bes subject. Every use of be thus has the properties of a raising verb.
The main difference between the three uses lies in the XPs VFORM value:
4
(31) a. copula be:
_
COMPS XP
_
b. progressive be:
_
COMPS VP[VFORM ing]
_
c. passive be:
_
COMPS VP[VFORM pass]
_
As given here, the copula be needs no further specication: any phrase that can function as a
predicate can be its COMPS value. The progressive be requires its complement to be VP[ing],
and the passive be requires its complement to be VP[pass]. Hence, examples like those in (32)
are straightforwardly generated:
(32) a. John is [
AP
happy about the outcome].
b. John was [
VP[ing]
seeing his children].
c. The children are [
VP[pass]
seen in the yard].
Auxiliary have is rather similar in its properties to auxiliary be, and it selects a past participle
VP complement.
(33) a. John has not sung a song.
b. Has John sung a song?
c. John hasnt been singing a song.
d. John has sung a song and Mary has , too.
Given facts like these, we can posit the following information in the lexical entry for auxiliary
have, part of the perfect aspect construction:
4
See Chapter 9 for the further discussion of passive constructions.
158
(34)
_
_
have
HEAD
_
_
POS verb
AUX +
_
_
SPR 1
COMPS
_
VP
_
_
VFORM en
SPR 1
_
_
_
_
_
The interaction of subcategorization and morphosyntactic information is enough to predict the
ordering restrictions among modals. For example, the auxiliaries have and be can follow a
modal since both have bse as its VFORM value:
(35) a. John can [
VP[bse]
have danced].
b. John can [
VP[bse]
be dancing].
In addition, we can predict the following ordering too:
(36) a. He has [seen his children].
b. He will [have [been [seeing his children]]].
c. He must [have [been [being interrogated by the police at that very moment]]].
(37) a. *Americans have [paying income tax ever since 1913].
b. *George has [went to America].
(37a) is ungrammatical since have requires a perfect participle VP. (37b) is out since the fol-
lowing VP is nite.
In some varieties of English, typically in British English, the main verb have also has the
specication [AUX +], as evidenced by the (b) examples below:
(38) a. You are a student.
b. You have not enough money.
(39) a. Are you a student?
b. Have you enough money?
The main verbs be and have show the NICE properties; even though they are main verbs, they
have the syntax of auxiliaries. This fact supports the idea that every sentence has a (main) verb
in it, at least, while the surface syntax of a verb is determined by whether it has the specication
[AUX +] or [AUX ].
8.3.3 Periphrastic do
Next we discuss the so-called dummy do, which is used as an auxiliary in the absence of other
(nite) auxiliaries. This do also exhibits the NICE properties:
(40) a. John does not like this town.
b. In no other circumstances does that distinction matter.
159
c. They didnt leave any food.
d. Jane likes these apples even more than Mary does .
Like the modals, do does not appear in nonnite clauses.
(41) a. *They expected us to do/should leave him.
b. I found myself needing/*doing need/*should needing sleep.
There are also some properties which distinguish do from other auxiliaries. First, unlike other
auxiliaries, do appears neither before nor after any other auxiliary:
(42) a. *He does be leaving.
b. *He does have been eating.
c. *They will do come.
Second, the verb do has no obvious intrinsic meaning to speak of. Except for carrying the
grammatical information about tense and agreement (in present tense), it has no semantic con-
tribution.
Third, if do is used in a positive statement, it needs to be emphatic (stressed). But in negative
statements and questions, no such requirement exists.
(43) a. *John does leave.
b. John DOES leave.
(44) a. John did not come.
b. John DID not come. (more likely in this case: John did NOT come.)
(45) a. Did John nd the solution?
b. How long did it last?
The most economical way of representing these lexical properties is to give do the lexical
entry given in (46).
(46)
_
_
do
HEAD
_
_
POS verb
AUX +
VFORM n
_
_
VAL
_
_
SPR 1 NP
COMPS
_
VP
_
_
AUX
VFORM bse
SPR 1
_
_
_
_
_
_
_
Like other auxiliaries including modals, do is specied to be [AUX +], which ensures that do is
sensitive to negation, inversion, contraction, and ellipsis (NICE properties), just like the other
160
auxiliaries. Further, do selects a subject NP and a VP complement whose unrealized subject is
structure-shared with its subject ( 1 ). Treating do as a raising verb like other English auxiliaries
is based on typical properties of raising verbs, one of which is that raising verbs allow expletives
as their subject, as we have seen above:
(47) a. John may leave.
b. It may rain.
c. *John may rain.
(48) a. John did not leave.
b. It did not rain.
c. *John did not rain.
The [AUX +] specication and raising-verb treatment of do account for the similarities to other
auxiliaries and modals.
The differences stemfrom the lexical specications on the feature values for HEAD|POS and
its complement VP. Unlike have and be, do is specied to be n. This property then accounts
for why no auxiliary element can precede do, for only the rst verb in a sequence may be nite.
(49) a. He might [have left].
b. *He might [do leave].
The requirement on the complement VP of the auxiliary do is [VFORM bse]. This feature
specication blocks modals from heading the VP following do, for modals are specied to be
[n], predicting the ungrammaticality of the examples in (50):
(50) a. *He does [can leave here].
b. *He does [may leave here].
The lexical entry further species that the complement of do is a VP[AUX]. This requirement
will correctly predict the ungrammaticality of examples in (51) and (52).
(51) a. *Jim [DOES [have supported the theory]].
b. *The proposal [DID [be endorsed by Clinton]].
(52) a. *I [do [not [have sung]]].
b. *I [do [not [be happy]]].
In (51) and (52), the VPs following the auxiliary do, stressed or not, bear the feature [AUX +]
inherited from the auxiliaries have and be. This explains their ungrammaticality.
5
5
There are special properties of do in imperatives, and different properties with dont. Do in imperatives can occur
before another auxiliary like be and have.
(i) a. Do be honest!
b. Dont be silly!
Do and dont in imperatives also have one distinct property: only dont allows the subject you to follow (try inserting
you in (ia) and (ib)). Their properties indicate that they have different lexical information from the verb forms used in
non-imperatives.
161
8.3.4 Innitival Clause Marker to
The auxiliary verbs to and do, in addition to differing by just one phonological feature, voicing,
differ in an important syntactic property: do appears only in nite contexts, and to only in non-
nite contexts. The verb to is, of course, the actual marker of the innitive in English. Even
though it has the form of a preposition, its syntactic behavior puts it in the class of auxiliary
verbs (cf. Gazdar et al. 1985):
(53) a. *John believed Kim to do not leave here.
b. John believes Kim not to leave here.
These verbs share the property that they obligatorily take bare verbal complements (hence, non-
base forms or modals cannot head the complement VP):
(54) a. *John believed Kim to leaving here.
b. *John did not leaving here.
c. *John expect to must leave.
d. *John did not may leave.
In terms of NICE properties, to also falls under the VP ellipsis criterion:
(55) a. Tom wanted to go home, but Peter didnt want to .
b. Lee voted for Bill because his father told him to .
These properties indicate that to should have a lexical entry like this:
(56)
_
_
HEAD
_
_
POS verb
AUX +
VFORM inf
_
_
SPR
_
1 NP
_
COMPS
_
VP
_
_
VFORM bse
SPR 1
_
_
_
_
_
It is an innitive auxiliary verb, whose complement must be headed by a V in the bse form.
8.4 Explaining the NICE Properties
In this section we discuss how we can account for the NICE properties, which are key diagnos-
tics for presence of auxiliary verbs.
8.4.1 Auxiliaries with Negation
The English negative adverb not leads a double life: one as a nonnite VP modier, marking
constituent negation, and the other as a complement of a nite auxiliary verb, marking sentential
negation. Constituent negation is the name for a construction where negation combines with
some constituent to its right, and negates exactly that constituent:
162
Constituent Negation: The properties of not as a nonnite VP modier can be supported
from its similarities with adverbs such as never in nonnite clauses as given in (57):
(57) a. Kim regrets [never/not [having seen the movie]].
b. We asked him [never/not [to try to call us again]].
c. Duty made them [never/not [miss the weekly meetings]].
Taking not to modify a nonnite VP, we can predict its various positional possibilities in non-
nite clauses, via the following lexical entry:
(58)
_
_
not
HEAD
_
_
POS adv
NEG +
MOD
_
VP[VFORM nonn]
_
_
_
_
_
The adverb not modies any nonnite VP:
(59) Constituent Negation:
VP[nonn]
.
.
.
.
.
.
.
.
.
.
.
.
.
Adv
_
MOD 1 VP[nonn]
_ 1 VP[nonn]
not . . . . . .
In the grammatical examples in (60) and (61), not modies a nonnite VP; in the ungrammatical
examples the VP[nonn] lexical constraint is violated.
(60) a. [Not [speaking English]] is a disadvantage.
b. *[Speaking not English] is a disadvantage.
c. *Lee likes not Kim.
(61) a. Lee is believed [not
VP[inf]
[to like Kim]].
b. Lee is believed to [not
VP[bse]
[like Kim].
c. *Lee is believed [to
VP[bse]
[like not Kim]].
Sentential Negation: In contrast to constituent negation, there is sentential negation, which
is the canonical expression of negation. One way to distinguish the two types of negation comes
from scope possibilities in an example like (62) (cf. Warner 2000).
(62) The president could not approve the bill.
Negation here could have the two different scope readings paraphrased in (63).
163
(63) a. It would be possible for the president not to approve the bill.
b. It would not be possible for the president to approve the bill.
The rst interpretation is constituent negation; the second is sentential negation.
Sentential not may not modify a nite VP:
(64) a. Lee never/*not left. (cf. Lee did not leave.)
b. Lee will never/not leave.
This construction shows one clear difference between never and not: not can only modify a
nonnite VP, a property further illustrated by the following examples:
(65) a. John could [not [leave the town]].
b. John wants [not [to leave the town]].
(66) a. *John [not [left the town]].
b. *John [not [could leave the town]].
Another difference between never and not is found in the VP ellipsis construction. Observe
the following contrast:
(67) a. Mary sang a song, but Lee never did .
b. *Mary sang a song, but Lee did never .
c. Mary sang a song, but Lee did not .
The data here indicate that not behaves differently from adverbs like never in nite contexts,
even though they all behave alike in nonnite contexts. never is a true diagnostic for a VP-
modier, and we use contrasts between it and not to reason what the properties of not must
be.
We have seen the lexical representation for constituent negation not above. Sentential not
appears linearly in the same position following a nite auxiliary verb but shows different
syntactic properties. The most economical way to differentiate sentential negation from con-
stituent negation is to assume that sentential negation is actually a syntactic complement of a
nite auxiliary verb (cf. Kim and Sag 1995, 2002). That is, we can assume that when not is
used as a marker of sentential negation, it is selected by the preceding nite auxiliary verb via a
lexical rule:
(68) Negative Auxiliary Verb Lexical Rule:
_
_
HEAD
_
_
AUX +
VFORM n
_
_
COMPS 1 XP
_
_
_
_
HEAD
_
_
AUX +
VFORM n
NEG +
_
_
COMPS Adv[NEG +], 1 XP
_
_
This lexical rule allows a nite auxiliary verb with a complement ( 1 ) to select an extra comple-
ment, marked [NEG +]. This rule gives a lexical entry which licenses the following structure
for sentential negation:
164
(69)
VP
_
_
VFORM n
AUX +
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
V
_
_
VFORM n
AUX +
COMPS 2 [NEG +], 3
_
_
2 Adv
3 VP[bse]
could
Adv
_
MOD 1 VP
_ 1 VP[VFORM bse]
_
VFORM n
NEG +
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
V
_
_
VFORM n
AUX +
NEG +
COMPS 2 [NEG +], 3 VP
_
_
2 Adv 3 VP[VFORM bse]
V S
.
.
.
.
.
.
.
Are NP VP
.
.
.
.
.
.
.
you V VP
.
.
.
.
.
.
.
.
studying English?
However, there are certain exceptions that present problems for the analysis of inverted auxil-
iaries involving a movement transformation. Observe the following contrast:
(74) a. I shall go downtown.
b. Shall I go downtown?
Here there is a semantic difference between the auxiliary verb shall in (74a) and the one in
(74b): the former conveys a sense of simple futurity in the near future, I will go downtown
whereas the latter example has a deontic sense, asking whether it is appropriate for me to go
downtown. If the inverted verb is simply moved from an initial medial position in (74b), it is
not clear how the grammar can represent this meaning difference.
English also has various interpretations for the subject-auxiliary inversion construction:
7
(75) a. Wish: May she live forever!
b. Matrix Polar Interrogative: Was I that stupid?
c. Negative Imperative: Dont you even touch that!
d. Subjunctive: Had they been here now, we wouldnt have this problem.
e. Exclamative: Boy, am I tired!
Each of these constructions has its own constraints which cannot fully be predicted from other
constructions. For example, in wish constructions, only the modal auxiliary may is possible.
In negative imperatives, only dont allows the subject to follow. These idiosyncratic properties
support a non-movement approach, in which auxiliaries can be specied to have particular uses
or meanings when inserted into particular positions in the syntax.
This in turn means that our grammar adds the following SAI grammar rule as a well-formed
condition:
7
See Fillmore (1999) for detailed discussion.
167
(76) Subject-Aux Inversion (SAI) Rule:
S
_
SPR
_
H
_
_
HEAD
_
_
INV +
AUX +
_
_
SPR A
COMPS B
_
_
, A, B
This rule thus licenses an an inverted, nite, auxiliary verb to combine with its subject (the SPR
value A) and complements (the COMPS value B ), forming a well-formed subject-auxiliary
inverted phrase like the following:
(77)
S
_
_
HEAD
_
_
AUX +
INV +
_
_
SPR
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
V
_
_
HEAD
_
_
AUX +
INV +
_
_
SPR 1 NP
COMPS 2 VP
_
_
1 NP 2 VP[VFORM bse]
_
PHON 1
HEAD
_
_
POS verb
VFORM n
AUX +
_
_
_
_
PHON 1 + nt
HEAD
_
_
VFORM n
AUX +
NEG +
_
_
_
_
This means that a word like can will be mapped to cant, gaining the NEG feature:
(82)
_
_
can
HEAD
_
_
POS verb
VFORM n
AUX +
_
_
_
_
cant
HEAD
_
_
VFORM n
AUX +
NEG +
_
_
_
_
As we have seen earlier, the head feature NEG will play an important role in forming tag ques-
tions:
(83) a. They can do it, cant they?
b. They cant do it, can they?
c. *They cant do it, cant they?
d. *They cant do it, can he?
The tag part of such a question has the opposite value for NEG compared to that in the main
part of the clause, and its subject needs to have the same index value as the matrix subject. For
this, we can introduce the feature XARG (external argument) for each predicate and link the
subject to this value.
8
This means that English has independently the following Tag Question
8
Traditionally, arguments are classied into external and internal ones in which the former usually refers to the
subject. The introduction of such a semantic feature is necessary if we want to make the subject value visible at the S
level. See Bender and Flickinger (1999) and Sag (2007).
169
Rule:
9
(84) Tag-Question Rule:
S S
_
_
NEG 1
XARG i
_
_
, S
_
_
NEG 1
INV +
XARG i
_
_
This rule means that a tag part can be added when it has the opposite NEG value whose subject
index is identical with that of the matrix subject. This will then project the following structure
for (83a):
(85)
S
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S
_
_
NEG +
XARG i
_
_
S
_
_
NEG
INV +
XARG i
_
_
.
.
.
.
.
.
.
.
.
.
.
.
1 NP
i
VP
_
_
NEG +
XARG i
SPR 1 NP
i
V
_
_
NEG
INV +
XARG i
SPR 2 NP
i
_
2 NP
They
V
_
_
NEG +
XARG i
SPR 1 NP
i
_
VP
can they
cant do it
As represented here, the NEG feature of the matrix verb is passed up to the rst S. The tag
question then needs to have the opposite NEG value in accordance with the rule in (84). The
semantic feature XARG identied with the subject starts from the auxiliary verb and then is
semantically composed into the meaning of S. The XARG value in a sense makes the subjects
9
This rule is a simplied version. To be more precise, we need to ensure that the second S, corresponding to the tag
part, must have only the auxiliary and the subject. See Bender and Flickinger (1999) and Sag (2007).
170
index value visible at the top level of the sentence in question so that the tag subject can also
refer to this.
8.4.4 Auxiliaries with Ellipsis
The standard generalization of Verb Phrase Ellipsis (VPE) is that it is possible only after an
auxiliary verb, as shown in the contrast (86) and (87).
(86) a. Kim can dance, and Sandy can , too.
b. Kim has danced, and Sandy has , too.
c. Kim was dancing, and Sandy was , too.
(87) a. *Kim considered joining the navy, but I never considered .
b. *Kim got arrested by the CIA, and Sandy got , also.
c. *Kim wanted to go and Sandy wanted , too.
The VP complement of an auxiliary verb, but not a main verb, can undergo VP ellipsis as long
as the context provides enough information for its interpretation.
The syntactic part of this generalization can be succinctly stated in the form of lexical rule:
(88) VP Ellipsis Rule:
_
_
HEAD| AUX +
COMPS XP
_
_
_
_
HEAD| AUX +
COMPS
_
_
As the rule is stated to apply to any XP after a verb with the [AUX +] specication, it can apply
to more than just VPs, and to more than just the canonical auxiliary verbs, but also be and have
in their main verb uses. With be, non-VP complements can be elided:
(89) a. Kim is happy and Sandy is too.
b. When Kim was in China, I was too.
The main verb have is somewhat restricted, but the contrast in (90) is clear. Even though have
is a main verb in (90a), it can allow an elided complement, unlike the main verb bring in (90b):
(90) a. A: Have you anything to share with the group?
B: No. Have you ?
b. A: Have you brought anything to share with the group?
B: No. *Have you bought ?
Given the lexical rule (88) which species no change in the ARG-ST, a canonical auxiliary
verb like can will have a counterpart minus its phrasal complement on the COMPS list:
(91)
_
_
can
SPR 1 NP
COMPS 2 VP[bse]
ARG-ST 1 , 2
_
_
can
SPR 1
COMPS
ARG-ST 1 , 2
_
_
171
Notice here that even though the VP complement is elided in the output, the ARG-ST is intact.
In the rst part of the example in (92), there are three auxiliary verbs:
(92) Kim must have been dancing and
_
_
a. Sandy must have been , too.
b. Sandy must have , too.
c. Sandy must , too.
_
_
There are therefore various options for an elided VP: the complement of been, or have, or must.
The analysis also immediately predicts that ellipsis is possible with the innitival marker to,
for this is an auxiliary verb, too:
(93) a. Tom wanted to go home, but Peter didnt want to .
b. Lee voted for Bill because his father told him to .
(94) a. Because John persuaded Sally to , he didnt have to talk to the reporters.
b. Mary likes to tour art galleries, but Bill hates to .
Finally, the analysis given here will also account for the contrast shown above in (67); a
similar contrast is found in the following examples:
(95) a. *Mary sang a song, but Lee could never .
b. Mary sang a song, but Lee could not .
The negator not in (95b) is a marker of sentential negation and can be the complement of the
nite auxiliary verb could. This means we can apply the VPE lexical rule to could after the
negation lexical rule (68), as shown (96):
(96)
_
_
could
SPR 1 NP
COMPS 2 Adv[NEG +], 3 VP[bse]
ARG-ST 1 , 2 , 3
_
_
could
SPR 1
COMPS 2
ARG-ST 1 , 2 , 3
_
_
As shown here in the right-hand form, the VP complement of the auxiliary verb could is not
realized as a COMP element, though the negative adverb is. This form would then project a
syntactic structure in (97):
172
(97)
VP
.
.
.
.
.
.
.
.
.
.
.
.
.
V
_
_
AUX +
SPR 1
COMPS 2 Adv[NEG +]
ARG-ST 1 , 2 , 3 VP[bse]
_
_
2 Adv
_
NEG +
_
could not
As represented here, the auxiliary verb could forms a well-formed head-complement structure
with not.
Why is there a contrast in the examples in (95)? The reason is that not can survive VPE
because it can be licensed in the syntax as a complement of an auxiliary, independent of the
following VP. However, an adverb like never is only licensed as a modier of VP (it is adjoined
to VP to give another VP), and hence if the VP were elided, we would have a hypothetical
structure like this:
(98)
VP
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
V[AUX +] *VP
could Adv[MOD VP]
never
Here, the adverb never modies a VP through the feature MOD, which guarantees that the
adverb selects the head VP that it modies. In an ellipsis structure, the absence of such a VP
means that there is no VP for the adverb to modify. In other words, there is no well-formed
phrasal structure predicting the ungrammaticality of *has never as opposed to has not.
10
10
As we have seen in 6.6.1, all modiers carry the head feature MOD whose value is the expression they modify.
173
8.5 Exercises
1. Each of the following sentences contains an item (in the parentheses) which we might
want to call an auxiliary. In each case, construct relevant examples that will clarify
whether it actually is an auxiliary:
(i) a. John got sent to prison. (got)
b. He ought to leave his luggage here. (ought)
c. They neednt take this exam. (need)
d. You better not leave it here. (better)
e. He dared not argue against his parents. (dared)
f. He used to go there very often. (used).
Explain your reasoning from the examples you provide.
2. Draw trees for the following sentences:
(i) a. The gardener must trim the rose bushes today.
b. This should be the beginning of a beautiful friendship.
c. I am removing the shovel from the shed.
d. The travelers have returned from their vacation.
e. Springeld would have built a police station with the federal grant.
f. Stringrays could have been cruising near the beach.
g. She seem to have given nancial assistance to an important French art
dealer.
3. Provide an analysis of the grammaticality or ungrammaticality of the following examples
together, with a tree structure for each, and lexical entries for the words playing the
crucial roles in the determination of grammaticality.
(i) a. Ann may spend/*spending/*spends/*spent her vacation in Italy.
b. It has rained/*raining/*rains/*rain every day for the last week.
c. Tagalog is spoken/*speak/*speaks/*spoke in the Philippines.
d. The roof is leaking/*leaked/*leaks/*leak.
(ii) a. *Americans have musted pay income tax ever since 1913.
b. *George is having lived in Toledo for thirty years.
c. *The house is been remodeling.
d. *Margaret has had already left.
e. *A medal was been given to the mayor by the sewer commissioner.
f. *Does John have gone to the library?
g. *John seems fond of ice cream, and Bill seems, too.
(iii) a. Sam may have been being interrogated by the FBI.
b. *Sam may have been being interrogating by the FBI.
c. *Sam may be had been interrogating by the FBI.
174
4. Analyze the following sentences providing tree structures and lexical entries for the
verbs:
(i) a. Have social problems made police work difcult?
b. The senator should not have forgotten the concerns of her constituents.
c. Tokyo has not loosened trade restrictions.
d. They love to play golf, but I do not.
e. Did the doctor prescribe aspirin?
f. George has spent a lot of money, hasnt he?
g. Sandy will read your reports, but Harold will not.
What grammar rules are needed for your structures? Are any lexical rules involved in
getting to the correct forms of the verbs?
5. English allows what is called negative inversion as illustrated in (ii):
(i) a. He can hardly believe that its already over.
b. I could have little known that more trouble was just around the corner.
c. I have never been spoken to so rudely!
(ii) a. [Hardly] was there any rain falling.
b. [Little] did I know that more trouble was just around the corner.
c. [Never] have I been spoken to so rudely!
(iii) a. He had hardly collected the papers on his desk, had he/*hadnt he?
b. He never achieved anything, did he/*didnt he?
Draw tree structures for the sentences (ii) and provide the lexical entries for hardly, little
and never. The examples in (iii) indicate that these adverbs all involve some kind of
negation in the sentence in which they appear. In addition, think of how your analysis
can account for the unacceptable examples in (iv):
(iv) a. As a statesman, he scarcely could do anything worth mentioning.
b. As a statesman, scarcely could he do anything worth mentioning.
c. *As a statesman, scarcely he could do anything worth mentioning.
6. Observe the following contrast and state a rule that can describe the usage of words like
any. Does your rule involve negative words like not or hardly if so, how? In addition,
construct examples replacing any with some and determine if there are any differences
between these two types of word.
(i) a. *Anyone isnt sleeping in my bed.
b. *Any zebras cant y.
c. *Anything hasnt happened to his optimism.
d. *Any of the citizens hardly ever say anything.
(ii) a. I didnt nd any bugs in my bed.
b. Nobody told them anything.
c. We never found any of the unicorns.
(iii) a. Never have I stolen from any members of your family.
175
b. Why havent any books been returned?
c. Hardly any of the citizens ever say anything.
7. After reading the following passage, provide lexical entries for the underlined words and
draw trees for the sentences which include them.
(i) This expanded role for auxiliaries in English has resulted in some curious
rules. One is that when a sentence is to be negated, the word not must
follow not the main verb (as used to be the case), but the auxiliary. This rule
creates an awkward dilemma in the occasional instance when the sentence
to be negated actually doesnt have an auxiliary verb. Thus, if I wish to deny
the sentence, I walked home, I must add an entirely meaningless auxiliary
from the verb do just to stand as the prop for the word not. The result is
the sentence, I didnt walk home. Now, do and did are often added to show
emphasis, but in those cases they are spoken with emphasis. Thus, there is
a difference between saying I didnt walk home and saying I DIDNT walk
home. The latter sentence expresses emphasis, but in the former sentence
the verb did expresses nothing at all; it is merely there to hang the not on.
If we tried to say, I walked not home, this would have an unacceptably odd
sound to it. It would, indeed, sound archaic. English literature is full of
such archaisms, since putting not after the main verb was still good usage
in the time of Shakespeare and a century or more later.
11
11
Adopted from Creationism & Darwinism, Politics & Economics by Kelley L. Ross.
176
9
Passive Constructions
9.1 Introduction
One important aspect of syntax is how to capture systematic relations between related construc-
tions. For example, the following two sentences are similar in meaning:
(1) a. One of Koreas most famous poets wrote these lines.
b. These lines were written by one of Koreas most famous poets.
We recognize (1b) as the passive counterpart of the active sentence (1a). These two sentences are
truth-conditionally similar: they both describe the event of writing the lines by one Korean poet.
The only difference involves grammatical functions: in the active voice (1a), one of Koreas most
famous poets is the subject, whereas in the passive voice (1b), these lines is the subject.
Observing these differences, the question that arises is: Why do we use different voices
for expressing or describing the same situation or proposition? It is generally accepted that
the passive construction is used for certain discourse-motivated reasons. For example, when
it is more important to draw our attention to the person or thing acted upon, we use passive.
Compare the following:
(2) a. Somebody apparently struck the unidentied victim during the early morning
hours.
b. The unidentied victim was apparently struck during the early morning hours.
We can easily see here that the passive in (2b) assigns more attention to the victim than the
active in (2a) does. In addition, when the actor in the situation is not important or specic, there
is often a preference to use the passive voice:
(3) a. Targets can be observed at any angle.
b. During the early evening, Saturn can be found in the north, while Jupiter rises in
the east.
Similarly, we use the passive voice in formal, scientic, or technical writing or reports to place
an emphasis or an objective presentation on the process or principle being described. For exam-
ple, compare the following pair:
177
(4) a. I poured 20cc of acid into the beaker.
b. About 20cc of acid was poured into the beaker.
It is clear that unlike the active sentence (4a), the passive sentence (4b) assigns a more objective
perspective to the process described.
In this chapter, leaving aside these discourse- or genre-motivated features of the use of pas-
sive constructions, we will look into the syntactic and semantic relationships between active
and passive as well as the properties of different passive constructions.
9.2 Relationships between Active and Passive
Consider the two canonical active and passive counterpart sentences:
(5) a. The executive committee approved the new policy.
b. The new policy was approved by the executive committee.
How do these construction types differ?
Grammatical Functions and Subcategorization: As briey noted earlier, one of the main
differences we can observe between (5a) and (5b) is that the passive sentence has a promoted
object: the new policy is the passive sentence subject, while the notional subject the executive
committee is realized, optionally, in a PP (headed by by). By denition, a transitive verb form
such as taken or chosen must have an object:
(6) a. John has taken Bill to the library.
b. John has chosen Bill for the position.
(7) a. *John has taken to the library.
b. *John has chosen for the position.
Yet, with the passive construction of such verbs, the object NP is not present in post-verbal
position, and must not be:
(8) a. *The guide has been taken John to the library.
b. *The department has been chosen John for the position.
(9) a. John has been taken to the library.
b. John has been chosen for the position.
The absence of the object in the passive is due to the fact that the argument that would have
been the object of the active verb has been promoted to be the subject in the passive.
Apart from the direct object, other subcategorization requirement stays unchanged in a pas-
sive form. For example, the active form handed in (10) requires an NP and a PP[to] as its
complements, and the passive handed in (11) still requires the PP complement:
(10) a. Pat handed a book to Chris.
b. *Pat handed to Chris.
c. *Pat handed a book.
178
(11) a. A book was handed to Chris (by Pat).
b. *A book was handed (by Pat).
Other Selectional Properties: The third important property, following from the fact that the
active verbs object is promoted to the passive constructions subject, is that other selectional
properties of the verb are preserved. For example, if the usual postverbal constituent should
be an expletive form like it, this requirement is on the subject in the passive. Compare the
following:
(12) a. They believe it/*Stephen to be easy to annoy Ben.
b. They believe there to be a dragon in the wood.
(13) a. It/*Stephen is believed to be easy to annoy Ben.
b. There is believed to be a dragon in the wood.
If the active complement is itself a clause, so must the subject of the passive verb be a clause:
(14) a. No one believes/suspects [that he is a fool].
b. [That he is a fool] is believed/suspected by no one.
Finally, if the postverbal constituent can be understood as part of an idiom, so can the subject
in the passive:
(15) a. They believe the cat to be out of the bag.
b. The cat is believed to be out of the bag.
We thus can conclude that the subject of the passive form is the argument which corresponds to
the object of the active.
Morpho-syntactic changes: In addition to changes in argument realization, the passive con-
struction requires the auxiliary verb be in conjunction with the the passive form of the verb (a
subtype of the en form, see 5.2.1). In addition to passive be italicized in the examples below,
there can be other auxiliary verbs, with the passive auxiliary last in the sequence:
(16) a. John drove the car. The car was driven.
b. John was driving the car. The car was being driven.
c. John will drive the car. The car will be driven.
d. John has driven the car. The car has been driven.
e. John has been driving the car. The car has been being driven.
f. John will have been driving the car. The car will have been being driven.
Semantics: In terms of meaning, as noted above, there is no change in the semantic role
assigned to the argument which is the subject in the passive. The agent argument of active verb
is expressed as an optional oblique argument of the PP headed by the preposition by in the
passive, or not at all:
179
(17) a. Pat handed Chris a note.
b. Chris was handed a note (by Pat).
(18) a. TV puts ideas into childrens heads.
b. Ideas are put into childrens heads (by TV).
The observations above mean that any grammar needs to capture the following basic properties
of passive:
.
Passive turns the active object into the passive subject;
.
Passive leaves other aspects of the COMPS value of the active verb unchanged;
.
Passive optionally allows the active subject to be the object in a PP headed by the preposition
by;
.
Passive makes the appropriate morphological change in the form of the main verb, and re-
quires the auxiliary be;
.
Passive leaves the semantics unchanged.
9.3 Approaches to Passive
There could be several ways to capture the systematic syntactic and semantic relationships
between active and passive forms. Given our discussion so far, one might think of relying on
grammatical categories in phrase structure (NP, VP, S, etc.), or on surface valence properties
(SPR and COMPS), often informally characterized as grammatical functions, or semantic roles
(agent, patient etc.). In what follows, we will see that we need to refer to all of these aspects of
the representation in a proper treatment of English passive constructions.
9.3.1 From Structural Description to Structural Change
Before we look into syntactic analyses for the formation of passive sentences, it is worth review-
ing Chomskys (1957) Passive Formation Rule formulated in terms of structural descriptions
(SD) and structural change (SC):
(19) Passive Formation Rule:
SD: X NP Y V NP Z
1
4
.
5
.
6
.
SC: 1 5 3 be 4+en 6 (by 2)
This rule means that if there is anything that ts the SD in (19), it will be changed into the given
SC: that is, if we have any string in the order of X NP Y V NP Z (in which X, Y,
and Z are variables), the order can be changed into X NP Y be V+en Z by NP. For
example, consider one example:
180
(20)
SD: Yesterday, the child really kicked a monkey in the street.
X NP Y V NP Z
1
4
.
5
6
.
SC: 1 5 3 be 4 + en 6 (by 2)
Yesterday, a monkey really was kicked in the street (by the child)
As noted here, the main change that occurs in the SC is that the rst NP became an optional
PP whereas the second NP became the rst NP. The rule also accompanies the addition of be
and the change of the main verbs VFORM into the passive form. Even though this old SD-SC
style rule does not reect constituenthood of the expressions in the given sentence and is not
satisfactory enough to account for all different types of passivization that we will see in the
following, this seminal work has inuenced the development of subsequent transformational
analyses for English passive constructions.
9.3.2 A Transformational Approach
A typical transformational approach assuming movement for passive involves the operation
shown in (21) (Chomsky 1982):
(21)
IP
.
.
.
.
.
.
.
NP
I
.
.
.
.
.
.
.
e I VP
.
.
.
.
.
.
.
Past V VP
.
.
.
.
.
.
.
be
V NP
deceived Bill
The operation moves the object Bill to the subject position and the verb be to I (In) position,
generating the output sentence Bill was deceived. This kind of movement analysis is based on
the following three basic assumptions:
.
Move : Move a category.
.
Case Theory: NP needs Case. The subject receives NOM (nominative) case from tense, and
the object receives ACC (accusative) from an active transitive verb.
1
1
In English, case is morphologically visible only on pronouns: he is nominative, whereas him is accusative, for
example.
181
.
A passive participle does not license ACC case.
In the lower position inside VP, the NP Bill in (21) cannot receive ACC case, since by as-
sumption the passive participle form deceived cannot assign any case. Hence, was deceived
Bill would violate Case Theory, for every NP must be assigned case. If the NP is moved to the
subject position, where case is assigned by the tensed verb was, then Case Theory is satised.
Even though this kind of movement or derivational analysis can be appealing in capturing the
relationships between canonical active and passive examples, it leaves many facts unexplained.
In what follows, we will see more complicated types of passive construction in English, and
the need to refer to not only grammatical categories but also grammatical functions, as well as
semantic/pragmatic constraints on passive.
9.3.3 A Lexicalist Approach
If we look into more passive examples, we can see that we need to refer to lexical and semantic
properties of the transitive verbs. First, there are many exceptions to passive. For example,
transitive verbs like resemble or t do not have any passive in some senses (see section 9.5
also):
(22) a. The model resembles Kim in nearly every detail.
b. *Kim is resembled by the model in nearly every detail.
(23) a. The coat does not t you.
b. *You are not tted by the coat.
Such a transitive verb presumably ts the tree structure in (21), but cannot be passivized.
In contrast, there are also verbs like rumor, say and repute which are used only in the passive,
as seen in the following contrast:
(24) a. I was born in 1970.
b. It is rumored that he is on his way out.
c. John is said to be rich.
d. He is reputed to be a good scholar.
(25) a. *My mother bore me in 1970.
b. *Everyone rumored that he was on his way out.
c. *They said him to be rich.
d. *They reputed him to be a good scholar.
Unlike verbs like resemble, these verbs are not used as active forms.
Such non-passive examples are hard to explain if we rely only on the assumption that passives
are derived from actives from congurational transformation rules. It seems that such lexical
idiosyncracies can be better treated in terms of a lexical process which allows us to refer to
the lexical and semantic properties of the verb in question. One way to capture these observed
lexical properties of passive is to posit a simplied lexical rule like the following:
182
(26) Passive Lexical Rule (to be revised):
_
_
trans-v
SPR 1 NP
COMPS 2 NP, . . .
_
_
_
_
pass-trans-v
HEAD| VFORM pass
SPR 2 NP
COMPS . . . ,
_
PP
_
by + 1 NP
_
_
_
This simple rule says that if there is a transitive verb (tran(sitive)-v) lexeme selecting one
SPR ( 1 ) and at least one COMPS element ( 2 and others), then there is a related passive verb
(pass-trans-v). This output verb selects the rst element on the original COMPS list as its SPR
( 2 NP) and the SPR in the input as an optional PP(consisting of by and 1 NP) with the remain-
ing COMPS value unchanged ( . . . ). The lexical process also accompanies the change in the
VFORM value into pass.
2
As it stands, this lexical rule is not precise enough. For example, consider the following:
(27) a. He kicked the ball. vs. b. The ball was kicked by him.
c. John kicked him. vs. d. He was kicked by John.
We can observe that the case on the rst argument of the predicate has changed from he in the
active to him in the passive. However, this difference is entirely predictable: all subjects of nite
clauses in English are nominative, and all objects of prepositions are accusative. So, rather than
stating the case changes directly, we can change the rule to refer only to the index value of the
subject and object:
3
2
As we noted in Chapter 5, in terms of the morphological form, the VFORM pass is a subtype of en.
3
Notice that the lexical rule given here can be represented in terms of the ARG-ST (Sag et al. 2002):
(i)
_
trans-v
ARG-ST XP
i
, XP
j
. . .
_
_
_
pass-trans-v
ARG-ST XP
j
, . . .
_
PP
i
[PFORM by]
_
_
_
This rule rearranges the elements of the input ARG-ST, also accompanying the change in the VFORM value into pass.
That is, the second element in the input ARG-ST becomes the rst element in the ARG-ST of the output passive verb.
Whatever follows the second argument in the input is intact, but the rst element in the ARG-ST becomes the object
of the optional PP. The Argument Realization Constraint in (4.4.3) will ensure that the rst element in the ARG-ST is
realized as the SPR and the remaining elements as the COMPS value as given in (28). Given that there is no discrepancy
between the ARG-ST values and the VAL (SPR and COMPS) values, we can formulate the passive rule as in (28) too.
183
(28) Passive Lexical Rule (Final):
_
_
trans-v
SPR XP
i
COMPS XP
j
, . . .
_
_
_
_
pass-v
HEAD| VFORM pass
SPR XP
j
COMPS . . . (PP
i
[PFORM by])
_
_
With this revised lexical rule, the case of the various NPs will be predicted by general prin-
ciples of case marking in English clauses, and need not be mentioned in the rule, which now
refers to the index values of the SPR and COMPS expressions. Let us see how all this works,
concentrating on a simple example:
(29) a. John sent her to Seoul.
b. She was sent to Seoul.
The active verb send is turned into the passive verb sent by the Passive Lexical Rule in (28):
(30) _
_
send
HEAD|POS verb
SPR NP
i
COMPS NP
j
, 2 PP[to]
_
_
sent
HEAD
_
_
POS verb
VFORM pass
_
_
SPR 1 NP
j
COMPS 2 , (PP
i
[by])
_
_
As seen here in the output form, the passive sent takes a SPR whose index value is identical
to that of the rst element of the COMPS list in the input. The passive sent also inherits the
PP[to] complement, tagged 2 , and selects an optional PP whose index value is identical to the
SPR (subject) of the input.
4
This output lexical entry will then license the following structure
for (29b):
4
As noted in Chapter 6.5.2, a preposition functioning as a marker rather than as a predicator with semantic content
does not contribute to the meaning of the head PP. This means that its index value is identical to that of its object NP.
184
(31)
S
_
VFORM n
_
_
_
_
_
_
_
_
_
_
_
_
_
2 NP
VP
_
_
VFORM n
SPR 2
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
N
V
_
_
VFORM n
SPR 2
COMPS 5
_
_
5 VP
_
_
VFORM pass
SPR 2
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
She was
V
_
_
VFORM pass
SPR 2
COMPS 3
_
_
3 PP
sent to Seoul
As given in the structure, the passive sent combines with its PP[to] complement, forming a VP
that still requires a SPR. This VP functions as the complement of the auxiliary be (was). As
we saw in Chapter 8, be is a raising verb, with the repeated lexical entry in (32), whose subject
(SPR value) is identical to its VP complements subject She:
(32)
_
_
be
HEAD| POS verb
SPR 2
COMPS
_
VP
_
_
VFORM pass
SPR 2
_
_
_
_
_
The SPR requirement on be is passed up to the highest VP in accordance with the VALP that
regulates the value of SPR and COMPS (see Chapter 5.1). When this VP combines with the
subject she in accordance with the Head-Specier Rule, the well-formed passive sentence is
complete.
The Passive Lexical Rule in (28) can be also applied to verbs which select for a CP comple-
ment. Consider the following examples:
(33) a. They widely believed that John was ill.
b. That John was ill was widely believed.
185
The application of the Passive Lexical Rule to the active believe will generate the passive output
shown in the following:
(34) _
_
believe
HEAD|POS verb
SPR NP
i
COMPS CP
j
_
believed
HEAD
_
_
POS verb
VFORM pass
_
_
SPR CP
j
COMPS (PP
i
)
_
_
The output passive verb believed then can license a structure like the following:
(35)
S
_
_
_
_
_
_
_
_
_
_
_
_
_
2 CP
VP
_
SPR 2
_
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
That John
was ill
V
_
SPR 2
COMPS 5
_
5 VP
_
VFORM pass
SPR 2
_
_
_
_
_
_
_
_
_
_
_
_
_
_
was Adv
VP
_
VFORM pass
SPR 2
_
.
.
.
.
.
.
.
.
.
.
.
.
.
widely
V
_
_
VFORM pass
SPR 2
COMPS( 3 )
_
_
3 PP
_
decide
HEAD|POS verb
SPR NP
i
COMPS S
j
[QUE +]
_
_
decided
HEAD
_
_
POS verb
VFORM pass
_
_
SPR S
j
[QUE +]
COMPS (PP
i
[by])
_
_
The output passive decided then will generate the following structure:
(38)
S
_
_
_
_
_
_
_
_
_
_
_
_
_
1 S[QUE +]
VP
_
SPR 1
_
Which attorney
will give the
closing
argument
V
_
SPR 1
COMPS 4
_
4 VP
_
SPR 1
_
has
V
_
SPR 1
COMPS 3
_
3 VP
_
SPR 1
_
been
V
_
_
VFORM pass
SPR 1 S[QUE +]
COMPS
_
_
decided
The passive verb decided selects an optional PP complement as its complement and an indirect
question as its subject. The raising verb rst combines with the rst VP, the result of which
again combines with the auxiliary raising verb has. Notice that since be and have are raising
verbs, their VP complement has the same subject as their own. By these identications, the
subject of has is identical to that of the passive verb decided.
5
We assume that indirect or direct questions are marked by the feature QUE(STION); see Chapter 10.
187
9.4 Prepositional Passives
In addition to the passivization of an active transitive verb, English also allows the so-called
prepositional verb to undergo passivization as illustrated in the following:
(39) a. You can rely on Ben.
b. Ben can be relied on.
(40) a. They talked about the scandal for days.
b. The scandal was talked about for days.
As we noted here, the object of the preposition in the active can function as the subject of the
passive sentence. Notice that such prepositional passives are possible with the verbs selecting a
PP with a specied preposition:
(41) a. The plan was approved of by my mother. (My mother approved of the plan.)
b. The issue was dealt with promptly. (They dealt with the issue promptly.)
c. Thats not whats asked for. (Thats not what they asked for.)
d. This should be attended to immediately. (We should attend to this immediately.)
(42) a. *Boston was own to. (They ew to/near/by Boston.)
b. *The capital was gathered near by a crowd of people. (A crowd of people gathered
near/at the capital.)
c. *The hot sun was played under by the children. (The children played under/near
the hot sun.)
The propositions in (41) are all selected by the main verbs (no other prepositions can replace
them). Meanwhile, each preposition in (42) is not selected by the main verb, since it can be
replaced by another one as noted in their active sentences.
One thing to observe is that there is a contrast between active and passive prepositional verbs
with respect to the appearance of an adverb (see Chomsky 1972, Bresnan 1982). Observe the
following:
(43) a. Thats something I would have paid twice for.
b. These are the books that we have gone most thoroughly over.
c. They look generally on John as selsh.
(44) a. *Everything was paid twice for.
b. *Your books were gone most thoroughly over.
c. *He is looked generally on as selsh.
The contrast here shows us that unlike the active, the passive does not allow any adverb to
intervene between the verb and the preposition.
There can be two possible structures that can capture these properties: ternary and reanalysis
structures. The ternary structure generates a at structure like the following:
188
(45)
VP[pass]
.
.
.
.
.
.
.
.
.
.
.
V[pass] P NP
V[pass]
.
.
.
.
.
.
.
.
.
.
.
NP
_
prep-v
SPR NP
i
COMPS PP
j
[PFORM 4 ]
_
_
pass-prep-v
VFORM pass
SPR NP
j
COMPS
_
P
_
_
LIGHT +
PFORM 4
_
_
, (PP
i
[PFORM by])
_
_
_
This rule ensures that a prepositional verb (prepositional-v) can have a counterpart passive
verb. This passive verb selects a SPR whose index value is identical to that of the input verbs
PP complement (in other words, the object of the preposition). The output passive verb also
has two complements: a preposition with the same PFORM as the input and an optional PP
complement expressing the agent argument (see below for the function of the feature LIGHT).
Lets see how the Prepositional Passive Rule and the Head-Light Rule combined together
can account for a prepositional passive:
(50) a. The lawyer looked into the document.
b. The document was looked into by the lawyer.
The active prepositional verb look can undergo the Prepositional Passive Lexical Rule as repre-
sented in the following:
(51)
_
_
look
SPR NP
i
COMPS PP
j
[into]
_
_
looked
VFORM pass
SPR NP
j
COMPS
_
P
_
_
LIGHT +
PFORM into
_
_
, (PP
i
[by])
_
_
_
The output passive verb selects one subject whose index value is identical to that of the inputs
PP complement. It also selects two complements: a preposition whose PFORM is identical with
that of the input PP and an optional PP[by] linked to the input subject. This output will then
license a structure like the following:
190
(52)
VP
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
V VP[pass]
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
was
V
_
_
VFORM pass
COMPS 3 PP
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3 PP
V
_
_
VFORM pass
COMPS 2 P, 3 PP
_
_
2 P
_
_
LIGHT +
PFORM into
_
_
by the lawyer
looked into
The Head-Light Rule in (48) allows the passive verb to combine with the preposition into rst,
forming still a lexical element. This resulting lexical element then combines with its PP com-
plement by the lawyer in accordance with the Head-Complement Rule.
8
8
This means that in the Head-Complement Rule the complement that the head combines with is phrasal.
191
9.5 Exercises
1. Draw tree structures for each of the following sentences and then provide the lexical
entry for the italicized passive verb.
(i) a. Peter has been asked to resign.
b. I assume the matter to have been led in the appropriate records.
c. Smith wants the picture to be removed from the ofce.
d. The events have been described well.
e. Over 120 different contaminants have been dumped into the river.
f. Heart disease is considered the leading cause of death in the United States.
g. The balloon is positioned in an area of blockage and is inated.
h. There was believed to have been a riot in the kitchen.
i. Cancer is now thought to be unlikely to be caused by hot dogs.
2. Provide the active counterpart of the following examples and explain how we can gener-
ate each of them, together with tree structures, lexical entries, and grammar rules.
(i) a. That we should call the police was suggested by her son.
b. Whether this is feasible hasnt yet been determined.
c. Paying taxes cant be avoided.
Also see if there are any relationships between the above examples and the following
passives:
(ii) a. It was suggested by her son that we should call the police.
b. It hasnt yet been determined whether this is feasible.
3. verbs like get and have can be used in so-called pseudo-passives:
(i) a. Frances has had the drapes cleaned.
b. Shirley seems to have Fred promoted.
(ii) a. Nina got Bill elected to the committee.
b. We got our car radio stolen twice on holiday.
In addition to these, have and get also allow constructions like the following:
(iii) a. Frances has had her clean the drapes.
b. Nina got them to elect Bill.
After drawing tree structures for the above examples, discuss the lexical properties of
have and get for example, what are their ARG-STs like?
4. Consider the following prepositional passive examples and then analyze them as far as
you can with tree structures.
(i) a. Ricky can be relied on.
b. The news was dealt with carefully.
c. The plaza was come into by many people.
d. The tree was looked after by Kim.
In addition, consider the passive examples in (ii):
192
(i) a. We cannot put up with the noise anymore.
b. He will keep up with their expectations.
(ii) a. This noise cannot be put up with.
b. Their expectations will be kept up with.
Can our analysis given in this chapter account for such examples? Also observe the
following examples (in (iv)), which illustrate two different kinds of passive:
(iii) a. They paid a lot of attention to the matter.
b. The son took care of his parents.
(iv) a. The matter was paid a lot of attention to.
b. A lot of attention was paid to the matter.
Can you think of any way to account for such examples?
5. We have seen that when the verb does not select a specied preposition, it usually does
not undergo passivization. However, observe the following contrast:
(i) a. *New York was slept in.
b. The bed was slept in.
(ii) a. *The lake was camped beside by my sister.
b. The lake is not to be camped beside by anybody.
Why do we have such a contrast with the same type of prepositional verb? In answering
this, think about the following contrast too, with respect to semantic or pragmatic facotrs:
(iii) a. *Six inches were grown by the boy.
b. *A pound was weighed by the book.
c. *A mile to work was run by him.
(iv) a. The beans were grown by the gardener.
b. The plums were weighed by the greengrocer.
In addition, can your semantic or pragmatic constraints explain the following contrast
too? If not, what kind of generalization can you think of to account the contrast here?
(v) a. *San Francisco has been lived in by my brother.
b. The house has been lived in by several famous personages.
(vi) a. *Seoul was slept in by the businessman last night.
b. This bed was surely slept in by a huge guy last night.
6. In certain environments, passives allow the auxiliary verb part to be get instead of be:
(i) a. Rosie got struck by lightning.
b. I got phoned by a woman friend.
c. He got hit in the face with the tip of a surfboard.
d. Johns bike got xed or got stolen.
Get passives usually convey the speakers personal involvement or reect the speakers
opinion as to whether the event described is perceived as having favorable or unfavor-
able consequences. This is why it is rather unacceptable to use the get passive when the
193
predicate is stative or the subject-referent has no control over the process in question:
(ii) a. *The king got feared by everyone.
b. *The lesson got read by a choirboy.
c. *The letter got written by a poet.
d. *Tom got understood to have asked for a refund.
e. *Mary got heard to insult her parents.
Based on these observations, provide the lexical entries for get in passive and tree struc-
tures for (ia) and (ib).
7. Read the following passages and identify all the grammatical errors in the verbs
VFORM values. In addition, provide the lexical information for the correct form.
(ii) Syntax is the discipline that examining the rules of a language that dictate
how the various parts of sentences gone together. While morphology looks
at how individual sounds formed into complete words, syntax looks at how
those words are put together for complete sentences. One part of syntax,
calling inection, deals with how the end of a word might changed to tell
a listener or reader something about the role that word is playing. Regular
verbs in English, for example, change their ending based for the tense the
verb is representing in a sentence, so that when we see Robert danced, we
know the sentence is in the past tense, and when we see Robert is dancing,
we know it is not. As another example, regular nouns in English become
plural simply by adding an s to the end. Cues like these play a large role
for helping hearers understanding sentences.
9
9
From http://www.wisegeek.com/what-is-syntax.htm
194
10
Wh-Questions
10.1 Clausal Types and Interrogatives
Like other languages, English also distinguishes a set of clause types that are characteristically
used to perform different kinds of speech acts:
(1) a. Declarative: John is clever.
b. Interrogative: Is John clever? Who is clever?
c. Exclamative: How clever you are!
d. Imperative: Be very clever.
Each clause type in general has its own functions to represent a speech act. For example, a
declarative makes a statement, an interrogative asks a question, an exclamative represents an
exclamatory statement, and an imperative issues a directive. However, these correspondences
are not always one to one. For example, the declarative in (2a) represents not a statement but a
question, while the interrogative in (2b) actually indicates a directive:
(2) a. I ask you if this is what you want.
b. Would you mind taking out the garbage?
In this chapter, we will focus on the syntactic structure of interrogatives, putting aside the map-
ping relationships between form and function.
There are basically two types of interrogative: yes-no questions and wh-questions:
(3) a. Yes-No questions: Can the child read the book?
b. Wh-questions: What can the child read?
Yes-no questions are different from their declarative counterparts by having subject and auxil-
iary verb in an inverted order. As we have seen in Chapter 8, such yes-no questions are generated
through the combination of an inverted nite auxiliary verb with its subject as well as with its
complement in accordance with the SAI Rule:
195
(4)
S
.
.
.
.
.
.
.
.
.
.
.
.
.
.
V
_
_
INV +
AUX +
SPR 1
COMPS 2
_
_
1 NP
2 VP
NP C
who C S
will NP
VP
they V VP
V NP
recommend t
The wh-phrase who is originally in the object position of recommend and then moved to the
specier position of the intermediate phrase C
NP S/NP
who V NP VP/NP
.
did they V NP/NP
.
recommend t
The notations such as NP/NP (read as NP slash NP) or S/NP (S slash NP) here mean that
the category to the left of the slash is incomplete and missing one NP. This missing information
is percolated up to the point where it meets its ller who. There is thus no notion of movement
here, but just a feature percolation up to the point where the missing gap meets its ller.
This kind of feature percolation system can account for the contrast given in (10a) and (10b).
Let us look at partial structures of these two examples:
198
(17) a.
S/NP
S/NP
and S/NP
NP VP/NP
NP VP/NP
S/NP
and S/PP
NP VP/NP
NP VP/PP
_
recommend
VAL
_
_
SPR 1
COMPS 2
GAP
_
_
_
_
_
_
verb-lxm
recommend
ARG-ST 1 , 2
_
b.
_
_
recommend
VAL
_
_
SPR 1
COMPS
GAP 2
_
_
_
_
In (22a), the two arguments of the verb are realized as the SPR and COMPS value respectively,
whereas in (22b) the second argument is realized not as a COMPS value but as a GAP value.
Each of these two different realizations will project the following structures for examples like
(19b) and (20b), respectively:
200
(23) a.
S
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
NP
VP[n]
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
The UN V[n] NP
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
recommended an enlarged peacekeeping force
b.
S[GAP ]
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
1 NP S[GAP 1 NP]
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
Who V NP VP[GAP 1 NP]
.
will they V[GAP 1 NP]
_
recommend
The main difference between the two is that in (23a), the object of recommend is its sister
whereas in (23b) it is not. That is, in the former the object is local to the verb whereas in the
latter it is nonlocal. In (23b), the verb contains a GAP value which is identied with the object.
This GAP value is passed up to the VP and then to the lower S. At this level, this GAP value is
discharged by the ller who, more specically, by the following Head-Filler Rule:
(24) Head-Filler Rule:
S
_
GAP
_
1 , S
_
GAP 1
_
This grammar rule says that when a head expression S containing a nonempty GAP value com-
bines with its ller value, the resulting phrase will form a grammatical head-ller phrase with
the GAP value discharged. This completes the top of the long-distance or unbounded depen-
dency.
10.3.2 Non-subject Wh-questions
Let us see how the present system generates a non-subject wh-question, using the verb put for
illustration. This verb will select three arguments as given here:
(25)
_
_
put
ARG-ST NP, NP, PP
_
_
The ARC will ensure that of these three arguments the rst one must be realized as a SPR
element, and the rest either as COMPS or as GAP elements. We thus will have at least the
201
following three realizations for the verb put:
1
(26)
a.
_
_
put
VAL
_
_
SPR 1
COMPS 2 , 3
GAP
_
_
ARG-ST
_
1 , 2 , 3
_
_
_
b.
_
_
put
VAL
_
_
SPR 1
COMPS 3
GAP 2
_
_
ARG-ST
_
1 , 2 , 3
_
_
_
c.
_
_
put
VAL
_
_
SPR 1
COMPS 2
GAP 3
_
_
ARG-ST
_
1 , 2 , 3
_
_
_
Each of these three lexical entries will then generate sentences like the following:
(27) a. John put the books in a box.
b. Which books did John put in the box?
c. Where did John put the books?
As we see here, the complements of the verb put are realized in three different ways. The
verb put in (26a) has the canonical realization of the verbs arguments, generating an example
like (27a). Meanwhile, in (26b), the object NP argument is realized as a GAP as reected in
(27b) whereas in (26c), the PP is realized as a GAP as shown in (27c). The following structure
represents how the lexical realization (26b) provides for examples like (27b) in more detail:
(28)
S[QUE +]
2 NP[QUE +]
S
_
GAP 2
_
Which book V
1 NP
VP
_
_
SPR 1
GAP 2
_
_
.
.
.
.
.
.
.
.
.
.
.
did John
V
_
_
SPR 1
COMPS 3
GAP 2
_
_
3 PP
1 NP S[GAP 1 NP]
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
do you V S[GAP 1 NP]
.
.
.
.
.
.
.
.
.
.
.
think NP VP[GAP 1 NP]
.
Hobbs V[GAP 1 NP]
_
met
The GAP value starts from the lexical head met whose second argument is realized as a GAP
value. Since the complement of the verb met is realized as a GAP value, the verb met will not
look for its complement in the local domain (as its sister node). The GAP information will be
passed up to the embedded S, which is a nonhead daughter. It is the following NIP that ensures
the GAP value in the head daughter or nonhead daughter to be passed up through the structure
until it is discharged by the ller who by the Head-Filler Rule:
3
(30) Nonlocal Feature Inheritance Principle (NIP):
A phrases nonlocal feature such as GAP, QUE, and REL is the union of its daugh-
ters nonlocal feature values minus any bound nonlocal features.
The role of this principle is clear from the embedded S in (29): The principle allows the GAP
in this nonhead S to pass up to the VP.
2
Every wh-element in questions carries the feature [QUE +].
3
For the feature REL, see Chapter 11.
203
With this principle together, we can observe that the treatment of long distance dependency
within the feature percolation system involves three parts: top, middle, and bottom. The bottom
part introduces the GAP value according to the ARC. The middle part ensures the GAP value
is inherited up to the mother in accordance with the NIP. Finally, the top level terminates the
GAP value by the ller in accordance with the Head-Filler Rule.
It is also easy to verify how this system accounts for examples like (31) in which the gap is
a non-NP:
(31) a. [In which box] did John put the book ?
b. [How happy] has John been ?
The Head-Filler Rule in (24) ensures that the categorial status of the ller is identical with that
of the gap. The structure of (31a) can be represented as following:
(32)
S
.
.
.
.
.
.
.
.
.
.
.
3 PP
S[GAP 3 PP]
.
.
.
.
.
.
.
.
.
.
.
did John
V
_
_
COMPS 2 NP
GAP 3 PP
_
_
2 NP
_
_
_
_
_
_
_
_
_
_
_
_
and
S
_
GAP 1 NP
_
_
_
_
_
_
_
_
_
_
_
_
_
NP
VP
_
GAP 1 NP
_
NP
VP
_
GAP 1 NP
_
_
_
_
_
_
_
_
_
_
_
_
_
and
S
_
GAP
_
_
_
_
_
_
_
_
_
_
_
_
_
NP
VP
_
GAP 1 NP
_
NP
VP
_
GAP
_
_
prep-v-p
on
ARG-ST 1 XP
_
_
b.
_
_
on
COMPS NP
GAP
_
_
c.
_
_
on
COMPS
GAP XP[nominal]
_
_
This realization means that when the preposition on is serving as the part of a prepositional
verb (prep-v-p) like rely on, its prepositional complement can be either realized as an NP or as a
nominal GAP element. This means when the argument of on is realized as a COMPS element,
it must be an NP. However, when its argument is realized as a GAP, the GAP value can be either
a CP or an NP since as we have seen in Chapter 5 type nominal subsumes both comp and noun.
This lexical realization will then project a structure like the following:
(38)
S
_
_
_
_
_
_
_
_
_
_
_
_
_
1 CP
S
_
GAP 1 [nominal]
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
you V
VP
_
GAP 1 [nominal]
_
can V
PP
_
GAP 1 [nominal]
_
rely
P
_
GAP 1 [nominal]
_
on
The present system, allowing certain exibility in the argument realization, thus can capture
movement paradox examples without resorting to any movement operations.
10.3.3 Subject Wh-Questions
Now consider examples in which the subject is the focus of the wh-question:
(39) a. Who put the book in the box?
206
b. Who DID put the book in the box?
c. Who can put the book in the box?
We can notice that when the subject who is questioned, the presence of an auxiliary verb is
optional. That is, the question in (39a) is well-formed, even though no auxiliary is present.
The related example (39b) is also well-formed, but is used only when there is emphasis on the
auxiliary.
As a rst step to account for such examples, we can, adopting a similar structure like non-
subject wh-questions, posit a structure like (40) in which the subject is gapped:
(40) a. Who put the book in the box?
b. Who can put the book in the box?
In the current context, our grammar requires no additional mechanism other than slightly
revising the ARC:
(41) Argument Realization Constraint (ARC, nal):
The rst element on the ARG-ST list is realized as SPR or GAP, the rest as
COMPS or GAP in syntax.
This revised ARC eventually guarantees that the values of the ARG-ST is the sum of that of
SPR, COMPS, and GAP. The system then allows the following lexical realization for put, in
addition to those in (26):
(42)
_
_
put
VAL
_
_
SPR
COMPS 2 , 3
GAP 1
_
_
ARG-ST
_
1 , 2 , 3
_
_
_
This realization in which the subject is gapped then projects the following structure for (40a):
207
(43)
S
_
_
SPR
COMPS
_
_
1 NP[QUE +]
S
_
_
SPR
COMPS
GAP 1
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Who
V
_
_
SPR
COMPS 2 , 3
GAP 1
ARG-ST 1 , 2 , 3
_
_
2 NP
3 PP
NP S[GAP 1 NP]
.
.
.
.
.
.
.
.
.
.
.
1 Who
V NP VP[GAP 1 NP]
.
.
.
.
.
.
.
.
.
.
.
do you V S[GAP 1 NP]
.
.
.
.
.
.
.
.
.
.
.
C S[GAP 2 NP]
Sara invited
4
The italicized hd-comps-ph and hd-mod-ph here indicates the type name of the phrase S.
209
b.
*CP
C
S[hd-comps-ph]
C S
_
_
hd-mod-ph
GAP 1 NP
_
_
that PP
S[GAP 1 NP]
_
wonder
HEAD| POS verb
SPR 1
COMPS
_
_
QUE +
_
_
_
_
b.
_
_
deny
HEAD| POS verb
SPR 1
COMPS
_
_
QUE
_
_
_
_
c.
_
_
tell
HEAD| POS verb
SPR 1
COMPS
_
_
QUE
_
_
_
_
The feature QUE originates from a wh-word like who or which and is used to distinguish be-
tween indirect questions and declarative clauses. The difference in the QUE value of the verbs
complement will ensure that each verb combines with an appropriate clausal complement. For
example, the verb wonder, requiring a [QUE +] clausal complement, will be licensed in a struc-
ture like the following:
211
(57)
VP
V
_
COMPS 2 [QUE +]
_ 2 S[QUE +]
wonder 1 NP[QUE +]
S[GAP 1 NP]
whose book NP
VP[GAP 1 NP]
his son V[GAP 1 NP]
likes
The GAP value of likes is passed up to the lower S and discharged by the ller whose book.
The wh-word whose carries the feature [QUE +] which, similar to the feature GAP, will pass up
to the point where it is required by a verb or to the highest position to indicate that the given
sentence is a question. For example, in (58), the feature QUE marks that the whole sentence is
a question, whereas in (59) it allows the embedded clause to be an indirect question:
(58) a. [
S[QUE +]
In which box did he put the book ]?
b. [
S[QUE +]
Which book by his father did he read ]?
(59) a. John asks [
S[QUE +]
in which box he put the book].
b. John asks [
S[QUE +]
which book by his father he read].
The percolation of the feature QUE upward from a wh-word can be ensured by the NIP that
guarantees nonlocal features like QUE to be passed up until they are bound off or selected by a
sister (like a ller phrase, or a selecting V). This principled constraint allows the QUE value to
pass up to the mother from a deeply embedded nonhead as illustrated in the following:
(60) a. Kim has wondered [[in which room] Gary stayed ].
b. Lee asked me [[how fond of chocolates] the monkeys are ]].
Let us consider the structure of (60a):
212
(61)
VP
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
V S[QUE +]
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
wondered
1 PP
_
QUE +
_
S[GAP 1 PP]
P
NP
_
QUE +
_
NP VP[GAP 1 PP]
in
Det
_
QUE +
_
N Gary V[GAP 1 PP]
which room stayed
Even though which is embedded in the PP and functions as the Det of the inner NP, its QUE
value will pass up to the S, granting it the status of an indirect question. The verb wonder then
combines with this S, satisfying its subcategorization requirement. If the verb combines with a
[QUE ] clausal complement, we would then have an ungrammatical structure:
(62) a. *Kim has wondered [
[QUE ]
that Gary stayed in the room].
b. *Kim asked me [
[QUE ]
that the monkeys are very fond of chocolates].
As we have seen above, the category of the missing phrase within the S must correspond to
that of the wh-phrase in the initial position. For example, the following structure is not licensed
simply because there is no Head-Filler Rule that allows the combination of the ller NP with an
S missing a PP:
213
(63)
VP
V *S[QUE +]
.
.
.
.
.
.
.
.
.
.
.
.
wondered NP[QUE +]
S[GAP 1 PP]
which room NP
VP[GAP 1 PP]
Gary V[GAP 1 PP]
stayed
In a similar fashion, the present system also predicts the following contrast:
(64) a. John knows [whose book [Mary bought ] and [Tom borrowed from her]].
b. *John knows [whose book [Mary bought ] and [Tom talked ]].
The partial structure of these can be represented as following:
(65) a.
S[QUE +]
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 NP[QUE +]
S[GAP 1 NP]
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
whose book
S[GAP 1 NP ]
and
S[GAP 1 NP ]
*S[GAP 1 NP]
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
whose book
S[GAP 1 NP ]
and
S[GAP 2 PP ]
_
whether
SYN
_
_
HEAD| POS comp
VAL| COMPS S
QUE +
_
_
_
_
According to this lexical information, whether selects a nite S and provides a [QUE +] value,
licensing a structure like the following:
(68)
CP[QUE +]
C[QUE +]
S[n]
VP[SPR NP[PRO]]
V
NP
V
VP
1 NP[QUE +]
S
_
_
VFORM inf
GAP 1
_
_
which politician
VP
_
_
VFORM inf
SPR PRO
GAP 1
_
to support
Consider the structure from the bottom up. The verb support selects two arguments whose
second argument can be realized as a GAP:
(78)
_
_
support
VAL
_
_
SPR 1 NP[PRO]
COMPS
GAP 2 NP
_
_
ARG-ST 1 NP, 2 NP
_
_
The verb will then form a VP with the innitival marker to. Since this VPs subject is PRO, the
VP can be projected into an S with the accusative NP GAP value in accordance with the Head-
Only Rule in (75). The S then forms a well-formed head-ller phrase with the ller which
politician. The QUE value on the phrase ensures the whole innitival clause to function as an
indirect question which can be combined with the verb knows.
One constraint we observe in the innitival wh-questions is that the innitival part cannot
have its subject to be realized:
(79) a. *Fred knows [which politician for Karen/her to vote for].
b. *Karen asked [where for Jerry/him to put the chairs].
218
The data indicate that in innitival indirect questions, the subject of the innitival VP cannot
appear. If we look at the structure, we can easily see why this is not a legitimate structure:
7
(80)
VP
.
.
.
.
.
.
.
.
.
.
.
.
.
V *S
_
QUE +
_
.
.
.
.
.
.
.
.
.
.
.
.
.
knows 1 NP[QUE + ]
CP
_
GAP 1 NP
_
.
.
.
.
.
.
.
.
.
.
.
.
.
which politician C S
_
GAP 1 NP
_
AdvP
_
MOD 1 S
_
1 S
_
_
AUX +
INV +
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
How carefully have you considered your future career?
The structure indicates that the AdvP modies the inverted S.
Matters become complicated when we consider questions with the interpretation of wh-
words (cf. Huang (1982)).
(83) a. When did he say that he was red?
b. Where did he tell you that he met Mary?
c. How did you guess that he xed the computer?
These sentences are ambiguous with respect to the local interpretation of the wh-words. (83a)
can question either the time he made the statement or the time he was red; (83b) can question
the place he told you the fact or the place he met Mary. The same is true in (83c): the wh-word
can question the proposition of the main clause or that of the embedded clause.
These data indicate that in addition to a structure like (82) in which the adverbial wh-word
modies the whole sentence, we need a structure where the adverbial wh-phrase linked to the
embedded clause. As a way of doing it, following Sag (2005) and others, we can assume that
English allows the extension of the ARG-ST to include a limited set of adverbial elements as
an argument. For example, we can extend the regular verb x to include a locative adverbial as
its argument.
(84) Extended ARG-ST:
_
_
x
ARG-ST 1 NP, 2 NP
_
_
_
_
x
ARG-ST 1 NP, 2 NP, AdvP
_
_
This extended ARG-ST then can allow us its adverbial argument to be realized as a GAP value
according to the ARC (Argument realization Constraint):
(85)
_
_
x
SPR 1 NP
COMPS 2 NP
GAP 3 AdvP
ARG-ST 1 NP, 2 NP, 3 AdvP
_
_
220
This lexical realization will then be able to project a structure like the following for the sentence
((83)c):
(86)
S
3 AdvP
S
_
GAP 3
_
How V NP
VP
_
GAP 3
_
did he V
S
_
GAP 3
_
guess NP
VP
_
GAP 3
_
he
V
_
GAP 3
_
NP
i
S
_
REL i
_
_
_
_
_
_
_
_
_
_
the 1 N
VP
_
MOD 1 N
_
.
.
.
.
.
.
.
.
.
boy
V
_
MOD 1 N
_
PP
1 N
i
, S
_
_
REL i
MOD 1
GAP
_
_
The rule, as a subtype of the Head-Modier Rule, basically ensures that a clause marked with
the REL feature modies a preceding noun with the identical index value. This grammar rule
now then ensures the presence of the MOD feature in the relative clause (8):
2
Following Sag (1997), one can develop an analysis in which the MOD value is introduced from a verb whose
argument contains a GAP value.
228
(15)
N
3 N
i
S
_
_
MOD 3 N
REL i
_
_
senators 2 NP[REL i]
S
_
GAP 2
_
whom NP
VP
_
GAP 2
_
Fred V
_
GAP 2
_
met
As shown here, the verb met realizes its object as a GAP value, which percolates up to the
S, where it is discharged by the relative pronoun whom. In addition, in accordance with the
Head-Rel Modier Rule in (14), the relative clause, forming a head-ller phrase, now carries
the MOD feature as well.
Since the relative clause is a type of head-ller phrase, there must be a total syntactic identity
between the gap and the ller with a REL value:
(16) a. Jack is the person [[
NP
whom] [Jenny fell in love with [
NP
]]].
b. Jack is the person [[
PP
with whom] [Jenny fell in love [
PP
]]].
(17) a. *Jack is the person [[
NP
whom] [Jenny fell in love [
PP
]]] .
b. *Jack is the person [[
PP
with whom] [Jenny fell in love with [
NP
]]].
In (16a) and (16b), the gap and the ller are the same category, whereas those in (17) are not.
The putative gap in (17a) is a PP and the one in (17b) an NP, but the llers are the non-matching
categories NP and PP, respectively.
In addition, the gap can be embedded in a deeper position as long as it nds the appropriate
ller:
229
(18)
N
_
_
_
_
_
_
_
_
_
_
_
_
_
2 N
i
S
_
_
REL i
MOD 2 N
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
video
1 NP
_
REL i
_
S
_
GAP 1 NP
_
_
_
_
_
_
_
_
_
_
_
_
_
_
which NP
VP
_
GAP 1 NP
_
I V
S
_
GAP 1 NP
_
thought NP
VP
_
GAP 1 NP
_
you recommended
The GAP value starts from the verb of the embedded clause and passes up to the top S in
accordance with the NIP. The value is discharged by the ller wh-phrase which carrying the
REL feature. This nonlocal REL feature, in accordance with the NIP, is also passed up to the
top S to ensure that the clause functions as a modier.
Just like the QUE feature, the nonlocal REL feature can also come from a deeper position
within the nonhead daughter of the relative clause:
(19) a. I met the critic [whose remarks [I wanted to object to ]].
b. This is the friend [for whose mother [Kim gave a party ]].
c. The teacher set us a problem[the answer to which [we can nd in the textbook]].
The simplied structure of (19b) can easily illustrate this point:
230
(20)
N
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
N
i
S[REL i]
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
friend
1 PP
_
REL i
_
S[GAP 1 PP]
P
NP
_
REL i
_
NP
VP[GAP 1 PP]
for
Det
_
REL i
_
N Kim gave a party
whose mother
The REL feature is embedded in the specier of the inner NP, but the NIP guarantees that this
value is passed up to the top S so that it can function as a modier to the antecedent friend.
11.3 Subject Relative Clauses
Subject relative clauses are not very much different from non-subject relatives clauses in terms
of modifying a nominal expression. One main difference is that the presence of a wh-relative
pronoun including that is obligatory, and bare relative clauses are ungrammatical:
(21) a. We called the senators [who] met Fred.
b. The kid picked up the apple [that] fell down on the ground.
(22) a. *[The student [ met John]] came.
b. *[The problem [ intrigued us]] bothered me.]
Subject relative clauses involve a missing subject a [REL i] subject is gapped, represented like
this:
231
(23)
N
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
3 N
i
S
_
_
REL i
MOD 3 N
i
GAP
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
senators
1 NP[REL i]
S
_
_
SPR
COMPS
GAP 1 NP
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
who
V
_
_
SPR
COMPS 2 NP
GAP 1 NP
ARG-ST 1 NP, 2 NP
_
_
2 NP
met Fred
As shown in the structure, the subject of met is realized as the GAP value which passes up to
the mother node. As noted in the previous chapter, this mother node is an S with the empty
COMPS and SPR value though it looks like a VP. It is an S with a gap in it, and this combines
with the ller who by the Head-Filler Rule. The resulting S is a complete one carrying the REL
and MOD specications that allows the resulting clause to modify senators in accordance with
the Head-Rel Modier Rule.
Notice that this analysis then does not license bare subject relatives given like (22). The VP
with the missing subject met John cannot carry the MOD feature at all even if it can function as
an S that can combine either with a wh-question phrase or a wh-relative phrase. However, the
analysis also predicts that the subject of an embedded clause can be gapped in sentences like
the following:
(24) a. He made a statement [which [
S
everyone thought [
S
was really interesting and
important]]].
232
b. They all agreed to include those matters [[
S
which [everyone believed [
S
had
been excluded from the Treaty]]]].
As we saw in Chapter 10, verbs like think and believe combine with a CP, an S, or even a S
with the subject being gapped:
(25)
N
i
_
_
_
_
_
_
_
_
_
_
_
_
_
3 N
i
S
_
_
REL i
MOD 3 N
i
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
statement
1 NP
_
REL i
_
S
_
GAP 1 NP
_
_
_
_
_
_
_
_
_
_
_
_
_
_
which NP
VP
_
GAP 1 NP
_
everyone V
S
_
GAP 1 NP
_
thought
V
_
GAP 1 NP
_
AP
was interesting . . .
The VP was interesting here forms an S with the subject being gapped. This S combines
with the verb thought, forming an incomplete VP with the GAP information. This GAP value
percolates up to the lower S and then is discharged by the relative pronoun which which induces
the MOD value to the relative clause so that it can modify the antecedent statement.
11.4 That-relative clauses
As noted earlier, that can be used either as a complementizer or as a relative pronoun:
(26) Complementizer that:
a. Mary knows that John was elected.
b. That John was elected surprised Frank.
233
c. Mary told Bill that John was elected.
(27) Relative Pronoun that:
a. This is the book [that we had read].
b. The president abandoned the people [that voted for him].
c. It is an argument [that people think will never end in Egypt].
Coordination data indicate that that can also be used a relative pronoun:
(28) a. *Every essay [shes written] and [that/which Ive read] is on that pile.
b. Every essay [which shes written] and [that Ive read] is on that pile.
c. Every essay [that shes written] and [which Ive read] is on that pile.
The contrast here can easily be accounted for if we assign the REL feature to that. As we have
seen earlier, the Coordination Rule requires two identical phrases to be conjoined. In (28b) and
(28c), two [REL i] Ss are conjoined, whereas in the unacceptable example (28a), two different
phrases, S with a gap element and S with no gap, are conjoined. This means that that will appear
in the following two different environments:
(29)
a. CP
Comp S
that . . .
b. S[REL i]
that . . .
The relative pronoun that is different from wh-relative pronoun in several respects. For ex-
ample, the relative pronoun that disallows genitive and piped piping (Sag (1997)):
(30) a. the student whose turn it was
b. *the student thats turn it was
(31) a. the pencil with which he is writing
b. *the pencil with that he is writing
In addition, that is used only in nite relative clauses:
(32) a. a pencil with which to write
b. *a pencil with that to write
One way to account for these differences from the other wh-relative pronouns is to assume
that the relative pronoun that has no accusative case, hence cannot be the complement of a
preposition that assigns accusative to it. The relative who, unlike relative pronouns like whose,
whom and which, has the same property in this respect:
(33) a. *The people [in who we placed our trust] . . . .
b. *The person [with who we were talking] . . . .
234
(34) a. The company [in which they have invested] . . .
b. The people [in whose house we stayed] . . .
c. The person [with whom he felt most comfortable] . . .
11.5 Innitival and Bare Relative Clauses
An innitival clause can also function as a modier to the preceding noun. Innitival relative
clauses in principle may contain a relative pronoun but need not:
(35) a. He bought a bench [on which to sit ].
b. He bought a refrigerator [in which to put the beer ].
(36) a. There is a book [(for you) to give to Alice].
b. There is a bench [(for you) to sit on].
Let us consider innitival wh-relatives rst. As we have seen in the previous chapter, an inni-
tival VP can be projected into an S when its subject is realized as the unrealized subject PRO.
This will then allow the following structure for (35a):
(37)
N
2 N
i
S
_
_
REL i
MOD 2 N
_
_
bench 1 PP[REL i]
S
_
GAP 1 PP
_
on which VP
_
GAP 1 PP
_
to sit
As shown here, the VP to sit has a GAP value which functions as the complement of sit. The
innitival VP, missing its PP complement, realizes its SPR as a PRO and thus can be projected
into an S in accordance with the Head-Only Rule. This S forms a head-ller phrase with the PP
on which. The resulting S also inherits the REL value from the relative pronoun which and thus
bears the MOD feature. Once again, we observe that every projection observes the grammar
rules as well as other general principles such as the HFP, the VALP, and the NIP.
Innitival wh-relatives have an additional constraint on the realization of the subject.
(38) a. a bench on which (*for Jerry) to sit
b. a refrigerator in which (*for you) to put the beer
235
The examples indicate that wh-innitival relatives cannot have an overt subject (such as (for)
Jerry) realized. We saw before that the same is true for innitival wh-questions; the data are
repeated here:
(39) a. Fred knows [which politician (*for Karen) to vote for].
b. Karen asked [where (*for Washington) to put the chairs].
This tells us that both innitival wh-relatives and innitival wh-questions are subject to the same
constraint. The reason for the ungammaticality of an example like (38a) can be understood if
we look at its structure:
(40)
N
.
.
.
.
.
.
.
.
.
.
.
.
.
*S
_
REL i
_
.
.
.
.
.
.
.
.
.
.
.
.
.
bench 1 PP[REL i]
CP
_
GAP 1 PP
_
.
.
.
.
.
.
.
.
.
.
.
.
.
on which C S
_
GAP 1 PP
_
Det N
[GAP ]
the 1 N
i
S
_
GAP NP
i
[acc]
_
paper
VP
_
_
SPR NP[PRO]
GAP NP
i
[acc]
_
_
to nish
The VP to nish has a GAP value for its object, and its subject is PRO. According to the Head-
only Rule in (75), this VP then will be projected into an incomplete S. There are two analytic
issues now: how to introduce the MOD feature and how to discharge the GAP value when there
is no ller. As we noted above, English also allows nite bare relatives with the gapped element
being accusative:
(43) a. the person [I met ]
b. the box [we put the books in ]
Note that unlike the traditional view, we can have bare relatives with the nominative subject
being gapped:
(44) a. He made a statement [everyone thought [ was interesting and important]].
b. They all agreed to include those matters [everyone believed [ had been excluded
from the Treaty]].
The subject gapped bare relative is only possible when it is combined by a matrix verb like
thought and believed, but not when it directly modies its antecedent as in (22). This in turn
means that we have the following constructional differences for bare relatives:
237
(45) a.
*N
S
_
GAP NP
_
S
_
GAP NP
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
statement everyone thought was interesting
As represented here, the simple S (VP with the subject being gapped) cannot function as a
bare relative modier, but the S whose embedded subject is gapped can well serve as a bare
relative clause. To reect this observation and allow us to discharge the GAP value in such bare
relative clauses, we can assume the following rule for English:
(46) Bare Head-Rel Modier Rule:
N
_
GAP
_
1 N
i
, S| CP
_
_
MOD 1
GAP NP
i
_
_
This rule allows a nite or innitival clause (S or CP, but not an S or VP) bearing an NP
GAP value to function as a modier of the preceding noun (the MOD value is added as a
constructional constraint). One specication in the rule is that the GAP value be discharged
even if there is no ller: the index of the head noun is identied with that of the discharged gap.
This rule will then allow examples like (44) as well as authentic examples like the following:
(47) I just know that the Big 12 South teams [everyone knew [ would win actually] won
the game].
11.6 Restrictive vs. Nonrestrictive Relative Clauses
In addition to the types of relative clause we seen before, there is an interpretive distinction
between restrictive and nonrestrictive. Consider these examples:
(48) a. The person who John asked for help thinks he is foolish.
b. Mary, who John asked for help, thinks he is foolish.
238
The relative clause in (48a) semantically restricts the denotation of person whereas the one in
(48b) just adds extra information about Mary. Let us consider one more pair of examples:
(49) a. John has two sisters who became lawyers. (restrictive)
b. John has two sisters, who became lawyers. (non-restrictive)
The second example suggests that John has only two sisters, while the rst means that two of
his sisters are lawyers, but leaves open the possibility that he has other sisters. The denotation
of the restrictive relative phrase two sisters who became lawyers is thus the intersection the set
of two sisters and the set of lawyers. There can be more than two sisters, but there are only two
who became lawyers. Meanwhile, the nonrestrictive phrase two sisters, who became lawyers
means that there are two sisters and they all became lawyers: there is no intersection meaning
here.
This meaning difference has given rise to the idea that the restrictive relative clause modies
the meaning of N
DP N
the N
NP
S
John, NP
whom we respect
Together with the differences in meaning, this structural difference can explain why the restric-
tive relative clause cannot modify a pronoun or proper noun:
3
3
In certain expressions of English, who relative clause can modify the pronoun he:
239
(52) a. I met the man who grows peaches.
b. I met the lady from France who grows peaches.
(53) a. *I met John who grows peaches.
b. *I met her who grows peaches.
Given that the meanings of John and her refer to only single individuals, we expect that no
further modication or restriction is possible. Nonrestrictive relative clauses as (54) can modify
proper nouns or pronouns, simply because they just add information about the referent into the
discourse:
(54) a. In the classroom, the teacher praised John, whom I also respect.
b. Reagan, whom the Republicans nominated in 1980, lived most of his life in Cali-
fornia.
The relative clause whom I also respect modies the proper noun John without restricting it,
and has the same interpretation as a conjoined clause like The teacher praised John, and I also
respect him.
Such a meaning difference also causes another difference: only a restrictive clause can mod-
ify a quantied NP like every N or no N:
(55) a. Every student who attended the party had a good time.
b. *Every student, who attended the party, had a good time.
(56) a. No student who scored 80 or more in the exam was ever failed.
b. *No student, who scored 80 or more in the exam, was ever failed.
Strictly speaking, phrases with no or every as determiners do not refer to an individual or given
set of individuals, and therefore cannot have their reference further elaborated by a nonrestric-
tive modier (see Huddleston and Pullum (2002)).
Whether the syntax involves N
NP
Det N
NP
.
.
.
.
.
.
.
.
.
.
.
.
.
NP
S
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
who won . . .
the contestant, who is the judges brother-in-law,
Only in (58a) can the rst relative clause be interpreted restrictively, as it is attached at the N
level. Strictly speaking, as noted above, (58b) is not ill-formed, but can only have an interpreta-
tion in which both relative clauses are nonrestrictive.
4
11.7 Constraints on the GAP
We have observed that in wh-interrogatives and relative clauses, the ller and the gap can be
in a long-distance relationship. Yet, there are constructions where this dependency seems to be
restricted in certain ways. Consider the following examples:
(59) a. [Who] did he believe [that he would one day meet ]?
b. [Which celebrity] did he mention [that he had run into ]?
(60) a. *[Who] did he believe [the claim that he had never met ]?
b. *[Which celebrity] did he mention [the fact that he had run into ]?
4
One additional difference between restrictive and nonrestrictive clauses is that that is used mainly in restrictive
clauses.
(i) a. The knife [which/that] he threw into the sea had a gold handle.
b. The knife, [which/??that] he threw into the sea had a gold handle.
241
Why do we have the contrast here? Let us compare the partial structures of (59a) with (60a):
(61) a.
VP[GAP NP]
V CP[GAP NP]
V *NP[GAP NP]
believe Det N
[GAP NP]
NP VP
He
i
V AP
is A CP
easy C IP
Op
i
NP VP
NP[PRO] V NP
please t
i
The subject he is base-generated in the matrix subject position, while the null operator Op
i
moves to the intermediate position from its original object position, leaving the trace (t
i
). At an
interpretive level, this operator is coindexed with the subject, indirectly linking the gap with the
ller even though the two have different case markings.
12.2.3 A Lexicalist Analysis
As we have seen earlier, unlike eager-type adjectives, easy-type adjectives require an incom-
plete VP complement as a lexical property. This subcategorization restriction appears to be a
lexical fact for a family of adjectives and verbs. In addition to adjectives like easy, verbs like
take and cost also select an innitival VP containing an accusative NP gap coindexed with the
subject:
(17) a. This theorem will take only ve minutes to prove .
b. This theorem will take only ve minutes to establish that he proved in 1930.
(18) a. This scratch will cost Kim $500 to x .
b. This $500 bribe will cost the government $500,000 to prove that Senator Jones
accepted .
Meanwhile, as we have noted in the previous section, eager-type adjectives do not have such a
subcategorization restriction.
We can represent this lexical difference in terms of lexical information. Let us begin with the
easy-type which selects a VP complement with one NP missing:
250
(19) easy-type adjectives
_
_
HEAD| POS adj
VAL
_
_
SPR NP
i
COMPS
_
_
_
VFORM inf
GAP 1 NP
i
[acc]
_
_
_
_
_
TO-BIND| GAP 1 NP
i
_
The lexical entry in (19) species that the innitival complement (VP or CP) of adjectives like
easy contains a GAP value (NP
i
) which is coindexed with the subject. This coindexation will
ensure the semantic linkage between the matrix subject and the gapped NP. Notice that unlike
canonical ller-gap constructions in which the GAP value is discharged when it meets the ller
(by the Head-Filler Rule), the feature TO-BIND is introduced to lexically discharge the GAP
value in the VP complement. This lexical information will then project the following structure
for (6a):
(20)
S
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
NP
i VP
_
_
_
_
_
_
_
_
_
_
_
_
Kim
V
AP
_
GAP
_
.
.
.
.
.
.
.
.
.
.
.
.
is
A
_
_
COMPS 2 VP[inf ]
TB-GAP 1 NP
_
_
2 VP
_
GAP 1 NP
_
easy V
VP
_
GAP 1
_
to
V
_
GAP 1 NP
i
[acc]
_
please
251
As shown in the tree, the transitive verb please introduces its object as the GAP value; hence the
mother innitival VP is incomplete. The adjective easy selects this VP and then lexically dis-
charges through the TO-BIND|GAP (TB-GAP) value in accordance with the following revised
NIP:
(21) Nonlocal Feature Inheritance Principle (NIP, nal):
A phrases nonlocal feature such as GAP and QUE is the union of its daughters
nonlocal feature values minus either the lexically or grammatically bound nonlo-
cal features.
Meanwhile, the lexical information for eager-type adjectives is very simple:
(22) eager-type adjectives
_
_
HEAD| POS adj
VAL
_
_
SPR NP
COMPS
_
VP
_
VFORM inf
_
_
_
_
_
_
These adjectives select a complete innitival VP with no missing element, eventually generating
a simple canonical head-complement structure like the following:
(23)
S
NP
VP
Kim
V AP
is
A
_
COMPS 2 VP
_
2 VP
_
GAP
_
eager V
VP
to V
please
The adjective eager thus places no restriction on its VP complement, and so can legitimately
combine with the fully saturated VP complement. When its VP complement has a GAP value,
252
it must be later discharged by a ller as seen in the following contrast:
(24) a. *Kim is eager to recommend .
b. Who is Kim eager to recommend ?
Notice that the present analysis can straightforwardly account for examples in which the VP
complement includes more than one GAP element. Compare the following pair of examples:
(25) a. This sonata is easy to play on this piano.
b. Which piano is this sonata easy to play on ?
The structure of (25a) is similar to that of (20):
(26)
S
.
.
.
.
.
.
.
.
.
.
.
.
1 NP
i
VP
_
SPR 1 NP
_
This sonata V
AP
_
_
SPR 1 NP
GAP
_
_
...........
_
_
_
_
_
_
_
_
_
_
_
_
is
A
_
_
SPR 1 NP
COMPS 3 VP
TB-GAP 2 NP
_
_
VP
_
_
SPR 1 NP
GAP 2 NP
_
_
_
_
_
_
_
_
_
_
_
_
_
_
easy V
3 VP
_
GAP 2 NP
i
_
.
.
.
.
.
.
.
.
.
S
_
GAP 2
_
Which piano V
1 NP
i
AP
_
_
SPR 1
GAP 2
_
_
............
is this sonata
A
_
_
SPR 1
COMPS 3
TB-GAP 4
_
_
3 VP
_
GAP 4 NP
i
, 2 NP
_
easy to play on
In the structure above, the VP complement of easy has two GAP values: one is the object ( 4 NP)
and the other is the object ( 2 ) of on. The rst GAP value coindexed with the subject is lexically
bound by easy through the feature TB-GAP. The remaining GAP value ( 2 NP) is passed up to
the second higher S where it is discharged by its ller, which piano, through the Head-Filler
Rule.
12.3 Extraposition
12.3.1 Basic Properties
English employs an extraposition process that places a heavy constituent such as a that-clause,
wh-clause, or innitival clause at the end of the sentence:
(28) a. [That dogs bark] annoys people.
b. It annoys people [that dogs bark].
(29) a. [Why she told him] is unclear.
b. It is unclear [why she told him].
(30) a. [(For you) to leave so soon] would be inconvenience.
b. It would be inconvenience [(for you) to leave so soon].
This kind of alternation is quite systematic: given sentences like (31a), English speakers have
an intuition that (31b) is possible:
(31) a. That the Dalai Lama claims Tibet independence discomts the Chinese govern-
ment.
254
b. It discomts the Chinese government that the Dalai Lama claims Tibet indepen-
dence.
The extraposition process can also be applicable to a clausal complement:
(32) a. I believe the problem to be obvious.
b. *I believe [that the problem is not easy] to be obvious.
c. I believe it to be obvious [that the problem is not easy].
As seen in (32bc), when a clausal complement is followed by innitival VP complement, the
former is much more preferably extraposed to sentence-nal position. In addition to a nite CP,
as in (32c), extraposition applies also to an innitival CP/VP, a simple S, or even a gerundive
phrase:
(33) a. I do not think it unreasonable [to ask for the return of my subscription].
b. He made it clear [he would continue to co-operate with the United Nations].
c. Theyre not nding it a stress [being in the same ofce].
12.3.2 Transformational Analysis
In terms of movement operations, there have been two main ideas to capture the systematic
relationships between examples such as the following:
(34) a. [That you came early] surprised me.
b. It surprised me [that you came early].
One approach assumes that the surface structure of a subject extraposition like (34b) is gener-
ated from (34a) as represented in the following (Rosenbaum (1967)):
3
(35)
S
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
NP
VP
It [t]
VP
V NP
you came early
surprised me
3
The notation [t] means a trace left after a movement.
255
The extraposition rule moves the underlying sentence you came early to a sentence-nal posi-
tion. This movement process also introduces the insertion of that, generating (34b). To generate
nonextraposed sentences like (34a) the system posits a process of deleting it in (34a) and then
adding the complementizer that.
A slightly different analysis has also been suggested with the opposite direction of movement
(Emonds 1970, Chomsky (1981), Groat (1995)). That is, instead of extraposing the clause from
the subject, the clause is assumed to be already in the extraposed position as in (36a):
(36) a. [[ ] [
VP
surprised [me] [
CP
that you came early]]].
b. [[It] [
VP
surprised me that you came early]].
The insertion of the expletive it in the subject position in (36a) would then account for (36b).
When the CP clause is moved to the subject position, the result is the nonextraposed sentence
(34a).
Most current movement approaches follow this second line of thought. Though such deriva-
tional analyses can capture certain aspects of English subject extraposition, they are not speci-
ed in enough detail to predict lexical idiosyncrasies as well as non-local properties of extrapo-
sition (see Kim and Sag (2005) for further discussion).
12.3.3 A Lexicalist Analysis
As we have seen, English exhibits a systematic alternation between pairs of non-extraposed and
extraposed sentences like the following:
(37) a. [That Chris knew the answer] occurred to Pat.
b. It occurred to Pat [that Chris knew the answer].
This alternation relation is quite productive. For example, as English acquires new expressions,
e.g. freak out, weird out, or bite, it acquires both extraposed and non-extraposed sentence types
(cf. Jackendoff 2002):
(38) a. It really freaks/weirds me out that we invaded Iraq.
b. That we invaded Iraq really freaks/weirds me out.
(39) a. It really bites that we invaded Iraq.
b. That we invaded Iraq really bites.
The simple generalization about the process of extraposition is that it applies to a verbal element
(CP, VP, and S). Adopting Sag et al. (2003), Kim and Sag (2006), we then can assume that the
extraposition process also refers to the verbal category whose subtypes include both comp and
verb (see Chapter 5.4.2). In particular, we can adopt the following lexical rule to capture the
systematic relationship in extraposition:
(40) Extraposition Lexical Rule (ELR):
_
ARG-ST . . . , 1 XP[verbal], . . .
_
_
_
ARG-ST . . . , NP[NFORM it], . . .
EXTRA 1 XP
_
_
256
What this rule says is that if a predicative element (actually, adjective or verb) selects a verbal
argument (either CP or S), this verbal element can be realized as the value of the feature EXTRA
together with the introduction of it as an additional argument.
For example, consider the following data set:
(41) a. Fidos barking annoys me.
b. That Fido barks annoys me.
c. It annoys me that Fido barks.
As shown here, the verb annoys can take either a CP or an NP as its subject. When the verb
annoys selects a verbal argument (CP), it can undergo the Extraposition Lexical Rule in (40) as
follows:
(42)
_
_
annoys
ARG-ST 1 [nominal] , 2 NP
_
_
_
_
ARG-ST NP[NFORM it], 2 NP
EXTRA 1 CP
_
_
Since the verb annoys selects a nominal (CP or NP since its subtypes are noun and comp) as
one of its arguments, it can undergo the ELR when this argument is realized as a CP (comp is a
subtype of verbal). As shown here, the output annoy now selects the expletive it as its subject
while its original CP now serves as the value of the EXTRA. The two arguments in the output
ARG-ST, in accordance with the ARC, will be realized as the SPR and COMPS value, with the
EXTRA value intact. This realization will allow us to generate a structure like the following:
257
(43)
S
3 NP
VP
_
_
SPR 3
EXTRA
_
_
It
VP
_
_
SPR 3
EXTRA 1
_
_
1 CP
V
_
_
SPR 3
COMPS 2
ARG-ST 3 NP[it], 2
EXTRA 1
_
_
2 NP
that Fido barks.
annoys me
As shown in the tree, the two arguments of the verb annoys are realized as SPR and COMPS
respectively. When it combines with the NP me, it forms a VP with a nonempty EXTRA value.
This VP then combines with the extraposed clause CP in accordance with the following Head-
Extra Rule:
(44) Head-Extra Rule:
_
EXTRA
_
H
_
EXTRA 1
_
, 1
As given here, the rule also discharges the feature EXTRA passed up to the head position.
This grammar rule reects the fact that English independently allows a phrase in which a head
element combines with an extraposed element:
(45)
_
EXTRA
_
H
_
EXTRA 1
_
1
258
We can observe that English freely employs this kind of well-formed phrase condition even in
the extraposition of an adjunct element.
(46) a. [[A man came into the room] [that no one knew]].
b. [[A man came into the room] [with blond hair]].
c. I [read a book during the vacation [which was written by Chomsky]].
All of these examples are licensed by the Head-Extra Rule which allows the combination of a
head element with an extraposed element.
Object extraposition is no different consider the following examples:
(47) a. Ray found the outcome frustrating.
b. Ray found it frustrating [that his policies made little impact on poverty].
The data indicate that the lexical entry for nd selects three arguments including a CP and thus
can undergo the ELR Rule:
(48)
_
_
nd
ARG-ST 1 NP, 2 [nominal], 3 AP
_
_
_
_
ARG-ST 1 NP, NP[it], 3 AP
EXTRA 2 [comp]
_
_
Since the type comp is a subtype of nominal and verbal at the same time, the verb can undergo
the ELR. The output introduces a new element it together with the EXTRA value. The three
arguments in the output will then be realized as the SPR and COMPS values, projecting a
structure like the following:
(49)
VP
_
EXTRA
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
VP
_
EXTRA 2
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2 CP
V
_
_
ARG-ST NP, 4 NP[it], 3 AP
EXTRA 2
_
_
4 NP 3 AP
that his policies . . .
found it frustrating
The verb nd requires an expletive object and an AP as its complement. It also has a clausal
element as its EXTRA element. The rst VP thus has a nonempty EXTRA value projected from
259
the verb, and this VP forms a well-formed phrase with the extraposed CP clause.
One main difference between subject and object extraposition is that the latter is obligatory:
(50) a. *I made [to settle the matter] my objective.
b. I made it [my objective] to settle the matter.
c. I made [the settlement of the matter] my objective.
(51) a. *I owe [that the jury acquitted me] to you.
b. I owe it [to you] that the jury acquitted me.
c. I owe [my acquittal] to you.
This contrast is due to a general constraint which disprefers any element within VP from occur-
ring after a CP:
(52) a. I believe strongly [that the world is round].
b. *I believe [that the world is round] strongly.
In the present context this means that there is no predicative expression (verbs and adjectives)
whose COMPS list contains an element on the list after a CP complement (see Kim and Sag
(2005)).
12.4 Cleft constructions
12.4.1 Basic Properties
The examples in (53) represent the canonical types of three kinds of cleft construction: it-cleft,
wh-cleft, and inverted wh-cleft in English:
(53) a. Its their teaching material that were using. (it-cleft)
b. What were using is their teaching material. (wh-cleft)
c. Their teaching material is what we are using. (inverted wh-cleft)
These three types of clefts all denote the same proposition as the following simple declarative
sentence:
(54) We are using their teaching material.
The immediate question that follows is then what is the extra function of the cleft struc-
ture instead of the simple sentence (54)? It is commonly accepted that clefts share identical
information-structure properties given in (55), for the example in question:
(55) a. Presupposition (Background): We are using X.
b. Highlighted (Foreground or focus): their teaching material
c. Assertion: X is their teaching material.
In terms of the structures, the three types of cleft all consist of a matrix clause headed by a copula
and a relative-like cleft clause whose relativized argument is coindexed with the predicative
argument of the copula. The only difference is where the highlighted (focused) expression is
placed.
260
12.4.2 Distributional Properties of the Three clefts
It-clefts: As noted before, the it-cleft construction consists of the pronoun it as the subject
of the matrix verb be, the highlighted (or focused) phrase XP, and a remaining cleft clause.
The pronoun it here functions as a place holder, though it is similar in form to the referential
pronoun it. For example, it is hard to claim that the pronoun it in the following dialogue has any
referential property:
(56) A: I share your view but I just wonder why you think thats good.
B: Well I suppose its the writer that gets you so involved.
As for the type of highlighted XP, we observe that only certain types of phrase can be used:
(57) a. It was [
NP
the man] that bought the articles from him.
b. It was [
AdvP
then] that he felt a sharp pain.
c. It was [
PP
to the student] that the teacher gave the best advice.
d. It was [
S
not until I was perhaps twenty-ve or thirty] that I read and enjoyed
them.
Phrases such as an innitival VP, AP, or CP cannot function as the XP:
(58) a. *It was [
VP
to nish the homework] that John tried.
b. *It is [
AP
fond of Bill] that John seems to be.
c. *It is [
CP
that Bill is honest] that John believes.
Also notice that in addition to that, wh-words like who and which can also introduce a cleft
clause:
(59) a. Its the second Monday [that] we get back from Easter holiday.
b. It was the girl [who] kicked the ball.
c. Its mainly his attitude [which] convinced the teacher.
Wh-clefts: Unlike the it-cleft, the wh-cleft construction places a cleft clause in the subject
position followed by the highlighted XP in the postcopular position. This gives a wide range of
highlighted phrases. As shown in (60), almost all the phrasal types can serve as the highlighted
XP:
(60) a. What you want is [
NP
a little greenhouse].
b. Whats actually happening in London at the moment is [
AP
immensely exciting].
c. What is to come is [
PP
in this document].
d. What Ive always tended to do is [
VP
to do my own stretches at home].
e. What I meant was [
CP
that you have done it really well].
Different from it-cleft, the wh-cleft allows AP, base VP, and clause (CP, simple S, and wh-
clause) to serve as the highlighted XP:
(61) a. What you do is [
VP
wear it like that].
261
b. What happened is [
S
they caught her without a license].
c. What the gentleman seemed to be asking is [
S
how policy would have differed].
Inverted wh-clefts: Though the inverted wh-cleft construction is similar to the wh-cleft, the
possible types of highlighted phrase are in fact different:
(62) a. [
NP
That] is what theyre trying to do.
b. [
AP
Insensitive] is how I would describe him.
c. [
PP
In the early morning] is when I do my best research.
(63) a. *[
VP
Wear it like that] is what you do.
b. *[
S
They caught her without a license] is what happened.
c. *[
CP
That you have done it really well] is what I meant.
In general, all wh-words except which are possible in inverted wh-clefts:
(64) a. Thats [when] I read.
b. That was [why] she looked so nice.
c. Thats [how] they do it.
d. Thats [who] I played with over Christmas.
e. *That was [which] I decided to buy.
12.4.3 Syntactic Structures of the Three Types of Cleft: Movement Analyses
There have been two main directions in movement analyses to deal with English it-cleft con-
structions: an extraposition analysis and an expletive analysis. The extraposition analysis as-
sumes a direct syntactic or semantic relation between the cleft pronoun it and the cleft clause
through extraposition (Akmajian 1970, Gundel 1977, Hedberg 2000).
(65) a. [What you heard] was an explosion. (wh-cleft)
b. It was an explosion, [what you heard]. (right-dislocated)
c. It was an explosion [that you heard]. (it-cleft)
For example, in Gundel (1977), the wh-cleft clause in (65a) is rst right dislocated as in (65b)
which then can generate the it-cleft (65c) with the replacement of what into that. Analyses of
this view basically take the cleft clause to be extraposed to the end of the sentence.
Meanwhile, the expletive analysis (Chomsky 1977,
E. Kiss 1998, Lambrecht 2001) takes the
pronoun it to be an expletive expression generated in place, while the cleft clause is semantically
linked to the clefted constituent by a predication relation.
(66) It was [
PRED
John + who heard an explosion].
An elaborated analysis within this view has been proposed by
E. Kiss (1998):
262
(67)
IP
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
NP I
.
.
.
.
.
.
.
.
.
.
.
.
.
It I FP
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
was
j
NP F
.
.
.
.
.
.
.
.
.
.
.
.
.
John
i
F CP
t
j
who
i
IP
.
.
.
.
.
.
.
.
.
.
.
t
i
heard an explosion
As shown here, the clefted phrase John, functioning as focus, is assumed to occupy the specier
of the FP (focus phrase) while the copula is the head of the FP and the cleft clause is the
complement of F. The highlighted focus phrase John and the cleft clause are thus in a predication
relation.
Even though the wh-cleft and it-cleft are identical in presenting salient discourse informa-
tion for emphasis, they have different syntactic properties which make it hard to derive one
from the other (cf. Pavey 2004). It is because there are many cases where we can observe clear
differences among the three types of clefts. For example, one noticeable difference lies in the
fact that only wh-clefts allow a base VP as the highlighted XP phrase:
(68) a. What you should do is [
VP
order one rst].
b. *It is [
VP
order one rst] that you should do rst.
c. *[
VP
Order one rst] is what you should do.
The three are different as well with respect to the occurrence of an adverbial subordinate clause:
(69) a. It was not until I was perhaps twenty-ve or thirty that I read them and enjoyed
them.
b. *When I read them and enjoyed them was not until I was perhaps twenty-ve.
c. *Not until I was perhaps twenty-ve was when I read them and enjoyed them.
As seen here, the not until adverbial clause appears only in it-clefts.
It is not difcult to nd imperfect relationships among the three types of cleft. For example,
neither wh-clefts nor inverted wh-clefts allow the cleft clause part to be headed by that:
(70) a. Its the writer [that gets you so involved].
b. *[That gets you so involved] is the writer.
263
c. *The writer is [that gets you so involved].
In addition, the head of the cleft clause in the it-cleft can be a PP, but not in the wh-cleft or
inverted wh-cleft:
(71) a. And it was this matter [[on which] I consulted with the chairman of the Select
Committee].
b. *[[On which] I consulted with the chairman of the Select Committee] was this
matter.
c. *This matter was [[on which] I consulted with the chairman of the Select Commit-
tee].
These facts suggest that the different types of cleft cannot be put in direct derivational rela-
tionships with each other. Though we cannot provide detailed analyses for them, we sketch out
possible directions here.
12.4.4 Lexically-Based Analyses
Wh-clefts: Let us rst consider wh-clefts:
(72) a. [What I ate] is an apple.
b. [What we are using] is their teaching material.
Before getting to the specic syntactic structures there are two things to note here: the role of
the copula be and the cleft clause. The copula in the cleft construction has a specicational
use, not a predicational one. In examples like (73a), the copula is predicational, whereas in
examples like (73b), the copula is specicational.
(73) a. The student who got A in the class was very happy.
b. The one who broke the window was Mr. Kim.
One main difference is that in the former the postcopular element denotes a property of the
subject whereas in the latter the postcopular element species the same individual as the subject.
In the wh-cleft too, the postcopular expression species the same individual as the subject.
As for the properties of the cleft part itself, we can observe that it behaves just like a free
relative clause. Not all wh-words can occur in free relatives:
(74) a. He got what he wanted.
b. He put the money where Lee told him to put it.
c. The concert started when the bell rang.
(75) a. *Lee wants to meet who Kim hired.
b. *Lee bought which car Kim wanted to sell to him.
c. *Lee solved the puzzle how Kim solved it.
In the examples in (74), what, where and when can head a free relative clause in the sense that
they are interpreted as the thing that, the place where, and the time when. However, this kind
264
of interpretation is not possible with who, which or how. As in free relatives, neither who nor
which can appear in wh-clefts, for example:
(76) a. *Who achieved the best result was Angela.
b. *Which book he read the book was that one.
Also note that the syntactic distribution of a free relative clause is as an NP, not as a clause of
some kind. For example, the object of eat is a diagnostic environment:
(77) a. I ate [what John ate].
b. I ate [an apple].
Since the verb ate requires only an NP as its complement, the only possible structure is as
follows:
(78)
S
NP
VP
I
V
NP
_
GAP
_
ate 1 NP
S
_
GAP 1
_
what
NP
VP
_
GAP 1
_
John
V
_
GAP 1 NP
_
ate
Even though the ller what and the head phrase John ate form a constituent, the result cannot be
an S since ate can combine with only an NP. This kind of free relative structure, rather unusual in
the sense that the non-head ller what is the syntactic head, is formed by the following grammar
265
rule (Pullum 1991):
4
(79) Free-Relative Phrase Rule:
NP[GAP ] 1 NP[FREL i], S[GAP 1 NP]
This rule ensures that when a free relative pronoun combines with a sentence missing one
phrase, the resulting expression is not an S but a complete NP.
On the assumption that the cleft clause in the wh-cleft is a free relative, we then can assign
the following structure to (72b):
(80)
S
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
NP
_
FREL i
_
VP
NP[FREL i] S/NP
V NP[FOC +]
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
NP
i
VP
_
_
_
_
_
_
_
_
_
_
_
_
_
Their teaching material V
NP
i
.
.
.
.
.
.
.
.
.
.
.
.
.
is
NP
S/NP
_
be
SPR NP[it]
COMPS
_
2 YP
i
[FOCUS +], CP
_
GAP 2
i
_
_
TO-BIND| GAP 2
i
_
(89) Copula be for Type B it-Cleft:
_
_
be
SPR NP[it]
COMPS
_
2 YP[FOCUS +], S
_
_
MOD 2
GAP
_
_
_
_
_
In both constructions, the contrastive focus (marked with the feature FOCUS functions as the
most salient contextual information. In Type A, the second complement, functioning as the
background, is a CP with a GAP value. The lexical head be also binds off the GAP inside the
CP complement. Notice that the index of the GAP value is identical to that of the focus YP,
providing a strict semantic linkage between these two.
In contrast, in Type B, the second COMPS element is a saturated S which modies the
focused element. Let us consider the structure that (88) licenses:
268
(90)
S
NP
VP
_
GAP
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
It
V
_
_
COMPS 2 , 3
TB-GAP 2
_
_
2 PP
i
_
FOCUS +
_
3 CP
_
GAP 2 PP
i
is on Bill C
S
_
GAP 2 PP
i
NP
VP
_
GAP
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
It
V
_
COMPS 2 , 3
_
2 NP
_
FOCUS +
_
3 S
_
_
GAP
MOD 2
_
_
is Tom
5 PP
S
_
GAP 5
_
NP
VP
_
GAP
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
It
V
_
COMPS 2 , 3
_
2 AdvP
_
FOCUS +
_
3 S
_
MOD 2
_
.
.
.
.
.
.
.
.
.
.
.
V
S
_
QUE +
_
.
.
.
.
.
.
.
.
.
.
.
.
.
wonder 2 NP i[QUE +]
S
_
GAP 2 NP i
_
.
.
.
.
.
.
.
.
.
.
.
.
.
who 1 NP VP
_
GAP 2 NP
_
.
.
.
.
.
.
.
.
.
.
.
.
.
it V
_
_
SPR 1
COMPS 3
GAP 2 NP
_
_
3 S
E. Kiss, Katalin. 1998. Identicational focus versus information focus. Language 74: 245273.
Emonds, Joseph. 1970. Root and structure-preserving transformations. Doctoral dissertation,
MIT.
Emonds, Joseph. 1975. A Transformational Approach to Syntax. New York: Academic Press.
Fillmore, Charles. 1963. The Position of Embedding Transformations in A Grammar. Word 19:
208-231.
Fillmore, Charles. 1999. Inversion and Constructional Inheritance. In G. Webelhuth, J.P Koenig,
and A. Kathol (eds.), Lexical and Constructional Aspects of Linguistics Explanation, 113
128. Stanford: CSLI Publications.
Fillmore, Charles J., Paul Kay, and Mary Catherine OConnor. 1988. Regularity and idiomatic-
ity in grammatical constructions: The case of let alone. Language 64(3): 501-538.
Fodor, Jerry A. 1983. The Modularity of Mind. Cambridge, MA: MIT Press
Fodor, Jerry A., and Jerrold J. Katz, (eds). 1964. The structure of language. Englewood Cliffs,
NJ: Prentice-Hall.
Fraser, Bruce. 1970. Idioms within a transformational grammar. Foundations of Language 6:
22-42.
Gazdar, Gerald. 1981. Unbounded dependencies and coordinate structure. Linguistic inquiry
12: 155-184.
Gazdar, Gerald. 1982. Phrase structure grammar. In P. Jacobson and G. K. Pullum (eds.), The
nature of Syntactic Representation. Dordrecht: Reidel.
Gazdar, Gerald, Ewan Klein, Geoffrey K. Pullum, and Ivan A. Sag. 1985. Generalized Phrase
Structure Grammar. Cambridge, MA; Havard University Press and Oxford; Basil Black-
278
well.
Gazdar, Gerald, and Geoffrey K. Pullum. 1981. Subcategorization, constituent order, and the
notion head. In M. Moortgat, H. van der Hulst, and T. Hoekstra (eds.), The Scope of
Lexical Rules. Dordrecht: Foris.
Gazdar, Gerald, Geoffrey K. Pullum, and Ivan A. Sag. 1982. Auxiliaries and related phenomena
in a restrictive theory of grammar. Language 58: 591-638.
Ginzburg, Jonathan, and Ivan A. Sag. 2000. Interrogative Investigations: The Form, Meaning
and Use of English Interrogatives. Stanford: CSLI Publications.
Goldberg, Adele E. 1995. A Construction Grammar Approach to Argument structure. Chicago:
University of Chicago Press. Green,
Grimshaw, Jane. 1997. Projection, Heads, and Optimality. Linguistic Inquiry, 28: 373422.
Green, Georgia M. 1976. Main Clause Phenomena in Subordinate Clause. Language 52: 382-
397.
Green, Georgia M. 1981. Pragmatics and Syntactic Description. Studies in the Linguistic Sci-
ences 11.1: 27-37.
Greenbaum, Sidney. 1996. The Oxford English Grammar. Oxford: Oxford University Press.
Groat, Erich M. 1995. English Expletives: A Minimalist Approach. Linguistic Inquiry 26.2:
354365.
Grosu, Alexander. 1972. The Strategic Content of Island Constraints. Ohio State University
Working Papers in Linguistics 13: 1-225.
Grosu, Alexander. 1974. On the Nature of the Left Branch Constraint. Linguistic Inquiry 5:
308-319.
Grosu, Alexander. 1975. On the Status of Positionally-Dened Constraints in Syntax. Theoret-
ical Linguistics 2: 159-201.
Grice, H. Paul. 1989. Studies in the Way of Words. Cambridge, MA: Harvard University Press.
Haegeman, Liliance. 1994. Introduction to Government and Binding Theory. Oxford and Cam-
bridge, MA: Basil Blackwell.
Harman, Gilbert. 1963. Generative grammar without transformation rules: A defense of phrase
structure. Language 39: 597-616.
Harris, Randy Allen. 1993. The Linguistic Wars. Oxford: Oxford University Press.
Harris, Zellig S. 1970. Papers in Structural and Transformational Linguistics. Dordrecht: Rei-
del.
Hooper, Joan, and Sandra Thompson. 1973. On the Applicability of Root Transformations.
Linguistic Inquiry 4: 465-497.
Hornstein, Norbert, and Daivd Lightfoot, (eds.) 1981. Explanation in linguistics: The logical
problem of language acquisition. London: Longman.
279
Huddleston, Rodney, and Geoffrey K. Pullum. 2002. The Cambridge Grammar of the English
Language. Cambridge University Press.
Hudson, Richard. 1984. Word Grammar. Oxford: Blackwell.
Hudson, Richard. 1990. English Word Grammar. Oxford: Blackwell.
Hudson, Richard. 1998. Word Grammar. In V. Agel, et al (eds.), Dependency and Valency: An
International Handbook of Contemporary Research. Berlin: Walter de Gruyter.
Huang, James. 1982. Logical Relations in Chinese and the Theory of Grammar. Ph.D. Disser-
tation, MIT.
Jackendoff, Ray. 1972. Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT
Press.
Jackendorff, Ray. 1975. Morphological and semantic regularities in the lexicon. Language 51:
639-671.
Jackendorff, Ray. 1994. Patterns in the Mind. New York: Basic Books.
Jackendoff, Ray. 1977. X
theory, 59
yes-no question, 195
292
This new textbook, focusing on the descriptive facts of English, provides a systematic introduc-
tion to English syntax for the students with no prior knowledge of English grammar or syntactic
analysis. The textbook aims to help students to appreciate the various sentence patterns avail-
able in English, understand insights into core data of English syntax, develop analytic abilities
to further explore the patterns of English, and learn precise ways of formalizing syntactic analy-
sis for a variety of English data and major English constructions such as agreement, raising and
control, the auxiliary system, passives, wh-questions, relative clauses, extraposition, and clefts.
Jong-Bok Kim is Associate Professor of School of English at Kyung Hee University, Seoul,
Korea. Peter Sells is Professor of Linguistics at Stanford University.
293